Up next in 10
Keynote: AI and GenAI for the Busy Software Architect | Software Architecture Conference
0 views
Aug 6, 2025
Staying ahead in an ever-evolving technological landscape is essential for any software architect. Explore fundamental concepts, practical applications, innovative strategies, and cutting-edge tools to adopt AI and GenAI technologies. Created for architects seeking to leverage AI for more intelligent, efficient, and future-proof systems, this keynote offers industry leaders and software architecture practitioners valuable knowledge and inspiration. š Conference Website: https://softwarearchitecture.live šŗ CSharp TV - Dev Streaming Destination http://csharp.tv š C# Corner - Community of Software and Data Developers https://www.c-sharpcorner.com #CSharpTV #CSharpCorner #CSharp #SoftwareArchitectureConf
View Video Transcript
0:03
awesome Hello friends and welcome to the
0:06
awesome Hello friends and welcome to the
0:06
awesome Hello friends and welcome to the Third Edition of the subware
0:08
Third Edition of the subware
0:08
Third Edition of the subware architecture
0:10
architecture
0:10
architecture conference and today is day two I hope
0:13
conference and today is day two I hope
0:14
conference and today is day two I hope you're enjoying the event since there
0:16
you're enjoying the event since there
0:16
you're enjoying the event since there are so many Fantastic sessions and
0:19
are so many Fantastic sessions and
0:19
are so many Fantastic sessions and technologies that are being shown in
0:22
technologies that are being shown in
0:22
technologies that are being shown in this
0:23
this
0:23
this conference and I'm Rodrigo Theon as I as
0:28
conference and I'm Rodrigo Theon as I as
0:28
conference and I'm Rodrigo Theon as I as the presentation that sign
0:31
the presentation that sign
0:31
the presentation that sign uh showed uh said and I'm a Microsoft
0:36
uh showed uh said and I'm a Microsoft
0:36
uh showed uh said and I'm a Microsoft Regional director I'm also a Microsoft
0:38
Regional director I'm also a Microsoft
0:39
Regional director I'm also a Microsoft MVP since 16 years ago currently in both
0:42
MVP since 16 years ago currently in both
0:42
MVP since 16 years ago currently in both the Microsoft vasher and development
0:45
the Microsoft vasher and development
0:45
the Microsoft vasher and development Technologies
0:47
Technologies
0:47
Technologies categories so we're truly living in
0:51
categories so we're truly living in
0:51
categories so we're truly living in exciting times right AI development has
0:55
exciting times right AI development has
0:55
exciting times right AI development has become more accessible especially with
0:57
become more accessible especially with
0:57
become more accessible especially with the rise of generative AI
1:00
the rise of generative AI
1:00
the rise of generative AI which is a type of artificial
1:03
which is a type of artificial
1:03
which is a type of artificial intelligence that generates text and
1:06
intelligence that generates text and
1:06
intelligence that generates text and also images in videos and audio and so
1:10
also images in videos and audio and so
1:10
also images in videos and audio and so many other outputs in response to
1:14
many other outputs in response to
1:14
many other outputs in response to prompts right for example here's a video
1:17
prompts right for example here's a video
1:17
prompts right for example here's a video that was shared the other day in X
1:21
that was shared the other day in X
1:21
that was shared the other day in X amazingly this person doesn't exist and
1:24
amazingly this person doesn't exist and
1:24
amazingly this person doesn't exist and it's becoming as you can see here it's
1:26
it's becoming as you can see here it's
1:26
it's becoming as you can see here it's becoming very difficult to identify
1:29
becoming very difficult to identify
1:30
becoming very difficult to identify what's real and what's not this video
1:33
what's real and what's not this video
1:33
what's real and what's not this video was created with the runway platform and
1:35
was created with the runway platform and
1:35
was created with the runway platform and one of its latest models you can go
1:38
one of its latest models you can go
1:38
one of its latest models you can go ahead and navigate to wrongway ml.com to
1:41
ahead and navigate to wrongway ml.com to
1:41
ahead and navigate to wrongway ml.com to see how you can create this fantastic
1:44
see how you can create this fantastic
1:44
see how you can create this fantastic kind of media assets so inspired by that
1:49
kind of media assets so inspired by that
1:49
kind of media assets so inspired by that video I open up the browser navigated to
1:52
video I open up the browser navigated to
1:52
video I open up the browser navigated to the wrongway web page and realize that
1:54
the wrongway web page and realize that
1:54
the wrongway web page and realize that you can also use a photo to create a a
1:59
you can also use a photo to create a a
1:59
you can also use a photo to create a a video from it right and here's the
2:01
video from it right and here's the
2:01
video from it right and here's the result you can see that I uploaded my
2:04
result you can see that I uploaded my
2:04
result you can see that I uploaded my picture and the video was created and
2:06
picture and the video was created and
2:06
picture and the video was created and it's kind of Well Creepy right kind of
2:10
it's kind of Well Creepy right kind of
2:11
it's kind of Well Creepy right kind of disturbing and you can also use the lip
2:14
disturbing and you can also use the lip
2:14
disturbing and you can also use the lip keyote Ai and geni for the busy software
2:16
keyote Ai and geni for the busy software
2:16
keyote Ai and geni for the busy software architect I hope you uh you uh listen to
2:21
architect I hope you uh you uh listen to
2:22
architect I hope you uh you uh listen to that audio you can also use the liping
2:24
that audio you can also use the liping
2:24
that audio you can also use the liping feature to input the text you want your
2:26
feature to input the text you want your
2:26
feature to input the text you want your photo to say so you can see that this is
2:30
photo to say so you can see that this is
2:30
photo to say so you can see that this is more disturbing right because now the
2:32
more disturbing right because now the
2:32
more disturbing right because now the photo is speaking this is this is
2:36
photo is speaking this is this is
2:36
photo is speaking this is this is amazing and you get the idea let's try
2:38
amazing and you get the idea let's try
2:38
amazing and you get the idea let's try this again welcome to my keynote Ai and
2:41
this again welcome to my keynote Ai and
2:41
this again welcome to my keynote Ai and G for the busy software so yeah I hope
2:45
G for the busy software so yeah I hope
2:45
G for the busy software so yeah I hope you're listening to the audio I'm not
2:47
you're listening to the audio I'm not
2:47
you're listening to the audio I'm not sure if the audio is being transmitted
2:49
sure if the audio is being transmitted
2:49
sure if the audio is being transmitted from my computer to your
2:52
from my computer to your
2:52
from my computer to your speakers and this is because I realized
2:55
speakers and this is because I realized
2:55
speakers and this is because I realized that I hadn't changed the voice that I
2:58
that I hadn't changed the voice that I
2:58
that I hadn't changed the voice that I selected beforehand and then I rendered
3:01
selected beforehand and then I rendered
3:01
selected beforehand and then I rendered the video with this other voice which is
3:03
the video with this other voice which is
3:03
the video with this other voice which is kind of okay let's try this again
3:06
kind of okay let's try this again
3:06
kind of okay let's try this again welcome to my
3:08
welcome to my
3:08
welcome to my K this is this is fantastic this is
3:11
K this is this is fantastic this is
3:11
K this is this is fantastic this is fascinating and this is frightening
3:14
fascinating and this is frightening
3:14
fascinating and this is frightening right in all seriousness we are truly
3:18
right in all seriousness we are truly
3:18
right in all seriousness we are truly living in exciting times right and of
3:21
living in exciting times right and of
3:21
living in exciting times right and of course media creation is just one
3:24
course media creation is just one
3:24
course media creation is just one example of the capabilities of
3:27
example of the capabilities of
3:27
example of the capabilities of generative AI according to Goldman Sachs
3:31
generative AI according to Goldman Sachs
3:31
generative AI according to Goldman Sachs generative AI could increase Global G
3:34
generative AI could increase Global G
3:34
generative AI could increase Global G GDP by as much as 7% or roughly s
3:40
GDP by as much as 7% or roughly s
3:40
GDP by as much as 7% or roughly s trillion dollars over the next 10 years
3:44
trillion dollars over the next 10 years
3:44
trillion dollars over the next 10 years which is fascinating right while many
3:47
which is fascinating right while many
3:47
which is fascinating right while many other predictions align with this there
3:50
other predictions align with this there
3:50
other predictions align with this there are some that they don't agree with this
3:54
are some that they don't agree with this
3:54
are some that they don't agree with this and that's fine the important takeaway
3:56
and that's fine the important takeaway
3:56
and that's fine the important takeaway is that generative AI is not going
3:59
is that generative AI is not going
3:59
is that generative AI is not going anyway is not going anywhere sorry and
4:02
anyway is not going anywhere sorry and
4:02
anyway is not going anywhere sorry and it's rapidly becoming a crucial part of
4:06
it's rapidly becoming a crucial part of
4:06
it's rapidly becoming a crucial part of everyone's lives okay and this is why
4:11
everyone's lives okay and this is why
4:11
everyone's lives okay and this is why software architecture is more important
4:14
software architecture is more important
4:14
software architecture is more important than ever especially if you're building
4:15
than ever especially if you're building
4:15
than ever especially if you're building software
4:17
software
4:17
software Solutions and you're planning to
4:20
Solutions and you're planning to
4:20
Solutions and you're planning to incorporate or adopt generative AI
4:24
incorporate or adopt generative AI
4:24
incorporate or adopt generative AI functionalities or features in your own
4:28
functionalities or features in your own
4:28
functionalities or features in your own applications so I think there are two
4:32
applications so I think there are two
4:32
applications so I think there are two primary ways to leverage generative AI
4:36
primary ways to leverage generative AI
4:36
primary ways to leverage generative AI enhancing and improving your own tasks
4:39
enhancing and improving your own tasks
4:39
enhancing and improving your own tasks in other in other
4:40
in other in other
4:40
in other in other words how you can perform
4:43
words how you can perform
4:43
words how you can perform better your tasks your daily tasks as a
4:47
better your tasks your daily tasks as a
4:47
better your tasks your daily tasks as a software architect and how you can
4:50
software architect and how you can
4:50
software architect and how you can improve your
4:52
improve your
4:52
improve your endusers tasks so let's see the first
4:56
endusers tasks so let's see the first
4:56
endusers tasks so let's see the first way as sare architect you don't work
4:59
way as sare architect you don't work
4:59
way as sare architect you don't work alone in an ivory Tower right you don't
5:02
alone in an ivory Tower right you don't
5:02
alone in an ivory Tower right you don't work in isolation I hope you don't do
5:05
work in isolation I hope you don't do
5:05
work in isolation I hope you don't do that so you're working alongside
5:08
that so you're working alongside
5:08
that so you're working alongside business professionals you're working
5:11
business professionals you're working
5:11
business professionals you're working alongside marketing team members
5:14
alongside marketing team members
5:14
alongside marketing team members software Engineers testers product
5:17
software Engineers testers product
5:17
software Engineers testers product managers and so on so those kind of
5:23
managers and so on so those kind of
5:23
managers and so on so those kind of processes and tasks and workflows when
5:26
processes and tasks and workflows when
5:26
processes and tasks and workflows when you want to communicate to those people
5:28
you want to communicate to those people
5:28
you want to communicate to those people you can improve those
5:30
you can improve those
5:30
you can improve those right so for instance with generative AI
5:34
right so for instance with generative AI
5:34
right so for instance with generative AI you can create documentation you can
5:36
you can create documentation you can
5:36
you can create documentation you can open up your favorite model and type
5:40
open up your favorite model and type
5:40
open up your favorite model and type your requirements or your code and then
5:44
your requirements or your code and then
5:44
your requirements or your code and then create the documentation for that
5:45
create the documentation for that
5:45
create the documentation for that particular project you can also create
5:49
particular project you can also create
5:49
particular project you can also create diagrams for instance you can create
5:52
diagrams for instance you can create
5:52
diagrams for instance you can create just one example is creating database
5:55
just one example is creating database
5:55
just one example is creating database diagrams you know relationships between
5:58
diagrams you know relationships between
5:58
diagrams you know relationships between tables and so on you can create those
6:01
tables and so on you can create those
6:01
tables and so on you can create those with generative AI Technologies you can
6:03
with generative AI Technologies you can
6:03
with generative AI Technologies you can also use generative AI for requirements
6:07
also use generative AI for requirements
6:07
also use generative AI for requirements analyses in other words you can input
6:10
analyses in other words you can input
6:10
analyses in other words you can input your requirements and let the model to
6:13
your requirements and let the model to
6:13
your requirements and let the model to give you some suggestions about the
6:17
give you some suggestions about the
6:17
give you some suggestions about the requirements and about the software
6:19
requirements and about the software
6:19
requirements and about the software architecture that you're trying to um
6:21
architecture that you're trying to um
6:21
architecture that you're trying to um that you're trying to create right
6:24
that you're trying to create right
6:24
that you're trying to create right tradeoff analysis this is another big
6:27
tradeoff analysis this is another big
6:27
tradeoff analysis this is another big one for us software Architects because
6:30
one for us software Architects because
6:30
one for us software Architects because you can input all the requirements and
6:33
you can input all the requirements and
6:33
you can input all the requirements and all the context and different scenarios
6:36
all the context and different scenarios
6:36
all the context and different scenarios and you can also grab a lot of different
6:39
and you can also grab a lot of different
6:39
and you can also grab a lot of different data sources like emails and um I don't
6:44
data sources like emails and um I don't
6:44
data sources like emails and um I don't know maybe uh documents in paper and you
6:47
know maybe uh documents in paper and you
6:47
know maybe uh documents in paper and you can scan those and grab the text and let
6:50
can scan those and grab the text and let
6:50
can scan those and grab the text and let the gener AI technology to give you an
6:53
the gener AI technology to give you an
6:53
the gener AI technology to give you an analysis of the tradeoffs about that
6:57
analysis of the tradeoffs about that
6:57
analysis of the tradeoffs about that particular project and you can also give
7:00
particular project and you can also give
7:00
particular project and you can also give you suggestions as I said and you can
7:02
you suggestions as I said and you can
7:02
you suggestions as I said and you can also automate processes right for
7:05
also automate processes right for
7:05
also automate processes right for instance maybe you are reporting to your
7:08
instance maybe you are reporting to your
7:08
instance maybe you are reporting to your team daily or each Friday or so on you
7:12
team daily or each Friday or so on you
7:12
team daily or each Friday or so on you can automate those processes by saying
7:14
can automate those processes by saying
7:14
can automate those processes by saying to the geni technology a you know what I
7:18
to the geni technology a you know what I
7:18
to the geni technology a you know what I want this report or this email to be
7:21
want this report or this email to be
7:21
want this report or this email to be sent out to my team daily or each Friday
7:27
sent out to my team daily or each Friday
7:27
sent out to my team daily or each Friday and so on right and it can also improve
7:32
and so on right and it can also improve
7:32
and so on right and it can also improve communication between team members uh as
7:34
communication between team members uh as
7:34
communication between team members uh as a software architect you're writing a
7:36
a software architect you're writing a
7:37
a software architect you're writing a lot of emails you're writing a lot of
7:39
lot of emails you're writing a lot of
7:39
lot of emails you're writing a lot of reports you're are communicating to
7:42
reports you're are communicating to
7:42
reports you're are communicating to those people that I mentioned before
7:44
those people that I mentioned before
7:44
those people that I mentioned before such as business
7:46
such as business
7:46
such as business professionals product managers marketing
7:49
professionals product managers marketing
7:49
professionals product managers marketing members and whatnot and you can use
7:53
members and whatnot and you can use
7:53
members and whatnot and you can use generative AI to create a template for
7:56
generative AI to create a template for
7:56
generative AI to create a template for those emails or those kind of uh
7:58
those emails or those kind of uh
7:59
those emails or those kind of uh communication
7:59
communication
7:59
communication mechanisms and let the geni technology
8:03
mechanisms and let the geni technology
8:03
mechanisms and let the geni technology to create the content and of course send
8:06
to create the content and of course send
8:06
to create the content and of course send it out to your team members so sofware
8:09
it out to your team members so sofware
8:09
it out to your team members so sofware AR architecture um tasks and workflow is
8:13
AR architecture um tasks and workflow is
8:13
AR architecture um tasks and workflow is just one example you can also use GNA
8:17
just one example you can also use GNA
8:17
just one example you can also use GNA for software development and software
8:19
for software development and software
8:19
for software development and software engineering practices such as code
8:22
engineering practices such as code
8:22
engineering practices such as code explanation this is a big one you can
8:25
explanation this is a big one you can
8:25
explanation this is a big one you can use today uh Technologies such as
8:28
use today uh Technologies such as
8:28
use today uh Technologies such as Copilot
8:30
Copilot
8:30
Copilot and let compilot to
8:32
and let compilot to
8:32
and let compilot to explain this particular code snippet
8:35
explain this particular code snippet
8:35
explain this particular code snippet that you maybe don't understand because
8:37
that you maybe don't understand because
8:37
that you maybe don't understand because you didn't create it or maybe you did
8:39
you didn't create it or maybe you did
8:40
you didn't create it or maybe you did and you uh time has passed and you don't
8:43
and you uh time has passed and you don't
8:43
and you uh time has passed and you don't remember what was the reasoning behind
8:45
remember what was the reasoning behind
8:45
remember what was the reasoning behind the code right that happens a lot and
8:49
the code right that happens a lot and
8:49
the code right that happens a lot and you can use copilot to explain the code
8:54
you can use copilot to explain the code
8:54
you can use copilot to explain the code and this is uh important for legacy code
8:58
and this is uh important for legacy code
8:58
and this is uh important for legacy code bases right
8:59
bases right
8:59
bases right us software developers and software
9:02
us software developers and software
9:02
us software developers and software Architects were
9:04
Architects were
9:04
Architects were struggling frequently with code bases
9:07
struggling frequently with code bases
9:07
struggling frequently with code bases that were created I don't know maybe 10
9:09
that were created I don't know maybe 10
9:09
that were created I don't know maybe 10 years ago or nine years ago with ancient
9:13
years ago or nine years ago with ancient
9:13
years ago or nine years ago with ancient Technologies such as I don't know maybe
9:16
Technologies such as I don't know maybe
9:16
Technologies such as I don't know maybe do NET Framework right or Visual Basic
9:21
do NET Framework right or Visual Basic
9:21
do NET Framework right or Visual Basic or an older version of java and so on so
9:26
or an older version of java and so on so
9:26
or an older version of java and so on so with this code explanation feature you
9:28
with this code explanation feature you
9:28
with this code explanation feature you can
9:29
can
9:30
can understand what's going on behind the
9:31
understand what's going on behind the
9:32
understand what's going on behind the reasoning of that particular code code
9:35
reasoning of that particular code code
9:35
reasoning of that particular code code generation of course you can create code
9:39
generation of course you can create code
9:39
generation of course you can create code and you can input the requirements this
9:41
and you can input the requirements this
9:41
and you can input the requirements this is uh uh you can use the copilot but of
9:45
is uh uh you can use the copilot but of
9:45
is uh uh you can use the copilot but of course you can use chat GPT or CLA or
9:49
course you can use chat GPT or CLA or
9:49
course you can use chat GPT or CLA or any other chat based jna platform you
9:54
any other chat based jna platform you
9:54
any other chat based jna platform you can input your requirements and let the
9:58
can input your requirements and let the
9:58
can input your requirements and let the GNA techn ol to Output the code for
10:01
GNA techn ol to Output the code for
10:01
GNA techn ol to Output the code for example you can say
10:03
example you can say
10:03
example you can say hey create code that communicates to
10:08
hey create code that communicates to
10:08
hey create code that communicates to This Server by using TCP sockets and the
10:13
This Server by using TCP sockets and the
10:13
This Server by using TCP sockets and the the I the address is this one and the
10:15
the I the address is this one and the
10:15
the I the address is this one and the port is this one and I want to use uh
10:18
port is this one and I want to use uh
10:19
port is this one and I want to use uh this class you can input those
10:22
this class you can input those
10:22
this class you can input those requirements and let the technology to
10:24
requirements and let the technology to
10:25
requirements and let the technology to Output the code and you can also say hey
10:27
Output the code and you can also say hey
10:27
Output the code and you can also say hey I want this code in Python I want this
10:29
I want this code in Python I want this
10:29
I want this code in Python I want this code this code to be created in
10:31
code this code to be created in
10:31
code this code to be created in JavaScript or C and whatnot right and
10:36
JavaScript or C and whatnot right and
10:36
JavaScript or C and whatnot right and code
10:38
code
10:38
code documentation you can grab your code
10:41
documentation you can grab your code
10:41
documentation you can grab your code your existing code and let the
10:43
your existing code and let the
10:43
your existing code and let the technology to create the documentation
10:44
technology to create the documentation
10:44
technology to create the documentation because
10:46
because
10:46
because nobody does documentation nowadays right
10:51
nobody does documentation nowadays right
10:51
nobody does documentation nowadays right um so we have this Technologies
10:53
um so we have this Technologies
10:53
um so we have this Technologies documentation is perfectly fine is uh
10:57
documentation is perfectly fine is uh
10:57
documentation is perfectly fine is uh quite needed and it's helpful
10:59
quite needed and it's helpful
10:59
quite needed and it's helpful the thing about documentation is that it
11:02
the thing about documentation is that it
11:02
the thing about documentation is that it becomes super outdated very rapidly
11:06
becomes super outdated very rapidly
11:06
becomes super outdated very rapidly right so with this kind of Technologies
11:09
right so with this kind of Technologies
11:09
right so with this kind of Technologies you can have your documentation to be in
11:12
you can have your documentation to be in
11:12
you can have your documentation to be in sync with your
11:15
sync with your
11:15
sync with your code and another example is creating
11:18
code and another example is creating
11:18
code and another example is creating synthetic data this is a big one when
11:21
synthetic data this is a big one when
11:21
synthetic data this is a big one when you are trying to
11:23
you are trying to
11:23
you are trying to create um some kind of data set that you
11:27
create um some kind of data set that you
11:27
create um some kind of data set that you want to use for say unit
11:30
want to use for say unit
11:30
want to use for say unit testing or for training your own models
11:34
testing or for training your own models
11:34
testing or for training your own models that's something that I truly don't
11:36
that's something that I truly don't
11:36
that's something that I truly don't recommend but if you're doing that maybe
11:39
recommend but if you're doing that maybe
11:39
recommend but if you're doing that maybe you need some kind of data set and you
11:41
you need some kind of data set and you
11:41
you need some kind of data set and you can create that data set by using
11:43
can create that data set by using
11:43
can create that data set by using synthetic data created by a large
11:46
synthetic data created by a large
11:46
synthetic data created by a large language model so I've used this in the
11:50
language model so I've used this in the
11:50
language model so I've used this in the past for unit testing this is perfectly
11:53
past for unit testing this is perfectly
11:53
past for unit testing this is perfectly uh
11:54
uh
11:54
uh acceptable uh mainly when you want to I
11:57
acceptable uh mainly when you want to I
11:58
acceptable uh mainly when you want to I don't know test some kind of function or
12:01
don't know test some kind of function or
12:01
don't know test some kind of function or or behavior that needs different
12:04
or behavior that needs different
12:04
or behavior that needs different variations of data and you don't have
12:07
variations of data and you don't have
12:07
variations of data and you don't have the enough imagination to create and and
12:10
the enough imagination to create and and
12:10
the enough imagination to create and and time of course to create that data okay
12:14
time of course to create that data okay
12:14
time of course to create that data okay so synthetic data creation is a uh
12:17
so synthetic data creation is a uh
12:17
so synthetic data creation is a uh another example that we can uh that we
12:21
another example that we can uh that we
12:21
another example that we can uh that we can say about uh ji and for product
12:25
can say about uh ji and for product
12:25
can say about uh ji and for product management as well so you can picture
12:29
management as well so you can picture
12:29
management as well so you can picture idea to this model and let the model
12:32
idea to this model and let the model
12:32
idea to this model and let the model give you suggestions or uh give you
12:35
give you suggestions or uh give you
12:35
give you suggestions or uh give you comments about your product idea about
12:38
comments about your product idea about
12:38
comments about your product idea about this application that you want that you
12:40
this application that you want that you
12:40
this application that you want that you want to create right for instance you
12:43
want to create right for instance you
12:43
want to create right for instance you can say hey I want to create this
12:45
can say hey I want to create this
12:45
can say hey I want to create this application that I don't know tracks the
12:50
application that I don't know tracks the
12:50
application that I don't know tracks the um activities and behavior of my kids in
12:52
um activities and behavior of my kids in
12:52
um activities and behavior of my kids in Comm
12:54
Comm
12:54
Comm kindergarten and I want to try this as
12:59
kindergarten and I want to try this as
12:59
kindergarten and I want to try this as scario and I want to create this mobile
13:01
scario and I want to create this mobile
13:01
scario and I want to create this mobile application and I want to do this and do
13:04
application and I want to do this and do
13:04
application and I want to do this and do that and let the technology give you
13:06
that and let the technology give you
13:06
that and let the technology give you back um suggestions and comments about
13:09
back um suggestions and comments about
13:09
back um suggestions and comments about that idea you can also create road maps
13:13
that idea you can also create road maps
13:13
that idea you can also create road maps you can create your uh product road map
13:17
you can create your uh product road map
13:17
you can create your uh product road map or application road map let the
13:19
or application road map let the
13:19
or application road map let the technology give you suggestions about
13:21
technology give you suggestions about
13:21
technology give you suggestions about that uh visual assets this is a big one
13:24
that uh visual assets this is a big one
13:24
that uh visual assets this is a big one you can see that I'm using some
13:27
you can see that I'm using some
13:27
you can see that I'm using some nicely uh done images that were created
13:31
nicely uh done images that were created
13:31
nicely uh done images that were created by using Bing uh image creation uh in
13:35
by using Bing uh image creation uh in
13:35
by using Bing uh image creation uh in the back end this is doll this is the
13:38
the back end this is doll this is the
13:38
the back end this is doll this is the technology that was created by open AI
13:41
technology that was created by open AI
13:42
technology that was created by open AI but you can see that you can create
13:44
but you can see that you can create
13:44
but you can see that you can create visual assets and you can also create
13:46
visual assets and you can also create
13:46
visual assets and you can also create text as I mentioned before and
13:49
text as I mentioned before and
13:49
text as I mentioned before and wireframes for your user interface
13:51
wireframes for your user interface
13:51
wireframes for your user interface that's another big one when you are
13:54
that's another big one when you are
13:54
that's another big one when you are trying to create something
13:56
trying to create something
13:56
trying to create something rapidly and you want to take advantage
14:00
rapidly and you want to take advantage
14:00
rapidly and you want to take advantage of jni for your products right so there
14:04
of jni for your products right so there
14:04
of jni for your products right so there are many other different scenarios where
14:06
are many other different scenarios where
14:06
are many other different scenarios where you can uh use jni when you are trying
14:10
you can uh use jni when you are trying
14:10
you can uh use jni when you are trying to improve or you're trying to um make
14:14
to improve or you're trying to um make
14:14
to improve or you're trying to um make more agile your your daily tasks right
14:18
more agile your your daily tasks right
14:18
more agile your your daily tasks right so what
14:19
so what
14:19
so what about this application that you're
14:21
about this application that you're
14:21
about this application that you're trying to create you're planning this
14:24
trying to create you're planning this
14:24
trying to create you're planning this solution and you want to implement gen
14:27
solution and you want to implement gen
14:27
solution and you want to implement gen AI Fe features and logic and uh behavior
14:32
AI Fe features and logic and uh behavior
14:32
AI Fe features and logic and uh behavior in your application what about that so
14:36
in your application what about that so
14:36
in your application what about that so here there are like thousands and
14:39
here there are like thousands and
14:39
here there are like thousands and thousands of options that we can uh use
14:43
thousands of options that we can uh use
14:43
thousands of options that we can uh use nowadays
14:45
nowadays
14:45
nowadays and uh most of the time those jna
14:49
and uh most of the time those jna
14:49
and uh most of the time those jna Solutions will be based on chatbots will
14:52
Solutions will be based on chatbots will
14:52
Solutions will be based on chatbots will be based on
14:54
be based on
14:54
be based on agents the difference between chatbots
14:56
agents the difference between chatbots
14:56
agents the difference between chatbots and agents is that a chat but you give a
15:00
and agents is that a chat but you give a
15:00
and agents is that a chat but you give a prompt or ask a question and it responds
15:04
prompt or ask a question and it responds
15:04
prompt or ask a question and it responds back with the uh with the response right
15:07
back with the uh with the response right
15:07
back with the uh with the response right with the with the answer that you're
15:08
with the with the answer that you're
15:08
with the with the answer that you're trying to find and agents are some kind
15:14
trying to find and agents are some kind
15:14
trying to find and agents are some kind of applications that are running all the
15:16
of applications that are running all the
15:16
of applications that are running all the time and agents instead of just
15:19
time and agents instead of just
15:19
time and agents instead of just responding back they do things okay they
15:24
responding back they do things okay they
15:24
responding back they do things okay they perform tasks this is this is important
15:30
perform tasks this is this is important
15:30
perform tasks this is this is important agents are Paramount and I believe
15:33
agents are Paramount and I believe
15:33
agents are Paramount and I believe agents are one of the most important uh
15:36
agents are one of the most important uh
15:36
agents are one of the most important uh and will be one of the most important um
15:39
and will be one of the most important um
15:39
and will be one of the most important um uh features ofi in the following months
15:43
uh features ofi in the following months
15:43
uh features ofi in the following months and years in other words you can create
15:46
and years in other words you can create
15:47
and years in other words you can create applications that do the work for you I
15:51
applications that do the work for you I
15:51
applications that do the work for you I was saying this example this scenario
15:53
was saying this example this scenario
15:53
was saying this example this scenario where you um send this report to your
15:57
where you um send this report to your
15:57
where you um send this report to your team members daily or each day or I
16:00
team members daily or each day or I
16:00
team members daily or each day or I don't know each Friday so of course that
16:04
don't know each Friday so of course that
16:04
don't know each Friday so of course that application needs access to your email
16:07
application needs access to your email
16:07
application needs access to your email needs access to the other people's uh uh
16:11
needs access to the other people's uh uh
16:11
needs access to the other people's uh uh email
16:12
email
16:13
email addresses and I don't know maybe it
16:15
addresses and I don't know maybe it
16:15
addresses and I don't know maybe it needs access to this particular database
16:18
needs access to this particular database
16:18
needs access to this particular database or documents since you want to you know
16:21
or documents since you want to you know
16:21
or documents since you want to you know create content based on that so agents
16:25
create content based on that so agents
16:25
create content based on that so agents agents my friends are what the what gen
16:29
agents my friends are what the what gen
16:29
agents my friends are what the what gen is going to be
16:31
is going to be
16:31
is going to be um where gen is going to be huge in the
16:35
um where gen is going to be huge in the
16:35
um where gen is going to be huge in the following following months and years of
16:38
following following months and years of
16:38
following following months and years of course you have to take advantage of
16:40
course you have to take advantage of
16:40
course you have to take advantage of multimodel models uh in other words
16:44
multimodel models uh in other words
16:44
multimodel models uh in other words models you can input text you can input
16:47
models you can input text you can input
16:47
models you can input text you can input images you can input videos and also
16:50
images you can input videos and also
16:50
images you can input videos and also audio uh for example you can see the uh
16:54
audio uh for example you can see the uh
16:54
audio uh for example you can see the uh open AI chat
16:55
open AI chat
16:55
open AI chat GPT application in your mobile and speak
17:00
GPT application in your mobile and speak
17:00
GPT application in your mobile and speak with the model right you're not sending
17:03
with the model right you're not sending
17:03
with the model right you're not sending text you're just speaking with this
17:06
text you're just speaking with this
17:06
text you're just speaking with this virtual thing that is responding back
17:10
virtual thing that is responding back
17:10
virtual thing that is responding back and with a n of course it comprehends
17:13
and with a n of course it comprehends
17:13
and with a n of course it comprehends natural language and you can speak as if
17:17
natural language and you can speak as if
17:17
natural language and you can speak as if you were speaking with other person and
17:19
you were speaking with other person and
17:19
you were speaking with other person and this application this technology this
17:22
this application this technology this
17:22
this application this technology this intelligence is responding back and
17:25
intelligence is responding back and
17:25
intelligence is responding back and through audio and of course video and
17:28
through audio and of course video and
17:28
through audio and of course video and images as well you can also create
17:31
images as well you can also create
17:31
images as well you can also create connecting the dots kind of
17:33
connecting the dots kind of
17:34
connecting the dots kind of solutions and this is where you want as
17:37
solutions and this is where you want as
17:37
solutions and this is where you want as I mentioned before you want to use data
17:40
I mentioned before you want to use data
17:40
I mentioned before you want to use data from your databases you want to use data
17:43
from your databases you want to use data
17:43
from your databases you want to use data from your emails right from your users
17:46
from your emails right from your users
17:46
from your emails right from your users inboxes or maybe you want to scan a lot
17:50
inboxes or maybe you want to scan a lot
17:50
inboxes or maybe you want to scan a lot of documents that you have in paper and
17:55
of documents that you have in paper and
17:55
of documents that you have in paper and extract some text from those documents
17:58
extract some text from those documents
17:58
extract some text from those documents and then let the intelligence connect
18:01
and then let the intelligence connect
18:01
and then let the intelligence connect the dots for you right for example hey I
18:05
the dots for you right for example hey I
18:05
the dots for you right for example hey I want to know what was the most important
18:07
want to know what was the most important
18:07
want to know what was the most important product that we have sold in the past
18:12
product that we have sold in the past
18:12
product that we have sold in the past three years or something right and let
18:15
three years or something right and let
18:15
three years or something right and let the intelligence connect the dots for
18:17
the intelligence connect the dots for
18:17
the intelligence connect the dots for you and of course media and content
18:20
you and of course media and content
18:20
you and of course media and content creation like the videos that I uh that
18:24
creation like the videos that I uh that
18:24
creation like the videos that I uh that I showed before right um and speak about
18:29
I showed before right um and speak about
18:29
I showed before right um and speak about those videos I I showed I HED you can
18:33
those videos I I showed I HED you can
18:33
those videos I I showed I HED you can sleep at night today after watching
18:35
sleep at night today after watching
18:35
sleep at night today after watching those videos okay so as a sofware
18:40
those videos okay so as a sofware
18:40
those videos okay so as a sofware architect you have to be aware of some
18:43
architect you have to be aware of some
18:43
architect you have to be aware of some quality attributes right software
18:45
quality attributes right software
18:45
quality attributes right software architecture is all about
18:48
architecture is all about
18:48
architecture is all about tradeoffs it's always about trade-offs
18:52
tradeoffs it's always about trade-offs
18:52
tradeoffs it's always about trade-offs so it's impossible to have all the
18:54
so it's impossible to have all the
18:54
so it's impossible to have all the quality attributes in the world in a
18:56
quality attributes in the world in a
18:56
quality attributes in the world in a single application right so you have to
19:01
single application right so you have to
19:01
single application right so you have to identify what are the most important
19:03
identify what are the most important
19:03
identify what are the most important quality attributes for your solution and
19:06
quality attributes for your solution and
19:06
quality attributes for your solution and when you're using gen and you want to
19:09
when you're using gen and you want to
19:09
when you're using gen and you want to adopt and uh uh Implement some kind of
19:13
adopt and uh uh Implement some kind of
19:13
adopt and uh uh Implement some kind of geni features in your Solutions security
19:16
geni features in your Solutions security
19:16
geni features in your Solutions security is Paramount okay security
19:20
is Paramount okay security
19:20
is Paramount okay security usability
19:22
usability
19:22
usability availability right because nobody wants
19:26
availability right because nobody wants
19:26
availability right because nobody wants this application this solution to to be
19:29
this application this solution to to be
19:29
this application this solution to to be available I don't know maybe just a
19:32
available I don't know maybe just a
19:32
available I don't know maybe just a couple of minutes each day right you
19:35
couple of minutes each day right you
19:35
couple of minutes each day right you have to have this availability for all
19:38
have to have this availability for all
19:38
have to have this availability for all your all your
19:40
your all your
19:40
your all your users and of course observability is a
19:44
users and of course observability is a
19:44
users and of course observability is a huge one as well since large language
19:48
huge one as well since large language
19:48
huge one as well since large language models and gen Technologies are
19:51
models and gen Technologies are
19:51
models and gen Technologies are stochastic in other words they you don't
19:54
stochastic in other words they you don't
19:54
stochastic in other words they you don't know uh if the next response will be the
19:58
know uh if the next response will be the
19:58
know uh if the next response will be the same as the previous one of course you
20:00
same as the previous one of course you
20:00
same as the previous one of course you can uh manage some parameters such as
20:03
can uh manage some parameters such as
20:03
can uh manage some parameters such as temperature and so on however the nature
20:07
temperature and so on however the nature
20:07
temperature and so on however the nature of
20:08
of
20:08
of jna is to be uh
20:12
jna is to be uh
20:12
jna is to be uh non-deterministic right so you have to
20:14
non-deterministic right so you have to
20:14
non-deterministic right so you have to be aware of that that's why
20:16
be aware of that that's why
20:16
be aware of that that's why observability is uh one of the most
20:19
observability is uh one of the most
20:19
observability is uh one of the most important quality attributes for your
20:21
important quality attributes for your
20:22
important quality attributes for your software Solutions when you're
20:23
software Solutions when you're
20:23
software Solutions when you're implementing this kind of solutions okay
20:27
implementing this kind of solutions okay
20:27
implementing this kind of solutions okay so security before I go to this other
20:30
so security before I go to this other
20:30
so security before I go to this other slide security is uh Paramount as well
20:33
slide security is uh Paramount as well
20:33
slide security is uh Paramount as well because
20:34
because
20:34
because jailbreaks this uh those are quite
20:37
jailbreaks this uh those are quite
20:37
jailbreaks this uh those are quite common nowadays when you're using large
20:40
common nowadays when you're using large
20:40
common nowadays when you're using large language models you don't want to do
20:42
language models you don't want to do
20:42
language models you don't want to do that and you don't want your users to
20:45
that and you don't want your users to
20:45
that and you don't want your users to jailbreak your jna Solutions okay so
20:50
jailbreak your jna Solutions okay so
20:50
jailbreak your jna Solutions okay so pattern Styles so patterns and styles
20:54
pattern Styles so patterns and styles
20:54
pattern Styles so patterns and styles speaking of uh software architecture of
20:56
speaking of uh software architecture of
20:56
speaking of uh software architecture of course you can use actually any kind of
21:00
course you can use actually any kind of
21:00
course you can use actually any kind of style or pattern that you're using
21:02
style or pattern that you're using
21:02
style or pattern that you're using nowadays you have to consider your model
21:07
nowadays you have to consider your model
21:07
nowadays you have to consider your model as any other module any other component
21:10
as any other module any other component
21:10
as any other module any other component in your overall architecture so I
21:14
in your overall architecture so I
21:14
in your overall architecture so I believe it's not different than having
21:16
believe it's not different than having
21:16
believe it's not different than having this I don't know this service this
21:19
this I don't know this service this
21:19
this I don't know this service this endpoint this component this I don't
21:22
endpoint this component this I don't
21:22
endpoint this component this I don't know uh you can use monolithic
21:24
know uh you can use monolithic
21:25
know uh you can use monolithic architectural style you can use
21:26
architectural style you can use
21:26
architectural style you can use microservices you can use serverless
21:29
microservices you can use serverless
21:29
microservices you can use serverless and you can also use and take advantage
21:32
and you can also use and take advantage
21:32
and you can also use and take advantage of edge Computing and this is going to
21:34
of edge Computing and this is going to
21:35
of edge Computing and this is going to be huge in the following months and
21:36
be huge in the following months and
21:36
be huge in the following months and years since
21:40
years since
21:40
years since small um language models are becoming uh
21:45
small um language models are becoming uh
21:45
small um language models are becoming uh more capable than ever so not large
21:49
more capable than ever so not large
21:49
more capable than ever so not large language models but small Lang large I'm
21:52
language models but small Lang large I'm
21:52
language models but small Lang large I'm sorry small language models that can be
21:55
sorry small language models that can be
21:55
sorry small language models that can be run in um small mobile phones or I don't
22:00
run in um small mobile phones or I don't
22:00
run in um small mobile phones or I don't know devices or laptops and tablets and
22:04
know devices or laptops and tablets and
22:04
know devices or laptops and tablets and so on okay so how you
22:08
so on okay so how you
22:08
so on okay so how you can uh transform your career path as an
22:12
can uh transform your career path as an
22:12
can uh transform your career path as an AI architect let's talk about this first
22:16
AI architect let's talk about this first
22:16
AI architect let's talk about this first skills you need to have problem solving
22:19
skills you need to have problem solving
22:19
skills you need to have problem solving and creativity okay you have to be
22:22
and creativity okay you have to be
22:22
and creativity okay you have to be creative you have to know how to talk to
22:25
creative you have to know how to talk to
22:25
creative you have to know how to talk to the model you have to know how to speak
22:28
the model you have to know how to speak
22:28
the model you have to know how to speak speak with this intelligence right I'm
22:30
speak with this intelligence right I'm
22:31
speak with this intelligence right I'm talking about prompt
22:33
talking about prompt
22:33
talking about prompt designing and prompting techniques okay
22:37
designing and prompting techniques okay
22:37
designing and prompting techniques okay and this is critical focus on the real
22:41
and this is critical focus on the real
22:41
and this is critical focus on the real Market needs okay you have to understand
22:44
Market needs okay you have to understand
22:44
Market needs okay you have to understand who the real customer is so there's
22:48
who the real customer is so there's
22:49
who the real customer is so there's no uh reason why you have you want to
22:53
no uh reason why you have you want to
22:53
no uh reason why you have you want to use Ai and gen just because you want to
22:56
use Ai and gen just because you want to
22:56
use Ai and gen just because you want to use it right don't uh fall into formal
23:01
use it right don't uh fall into formal
23:01
use it right don't uh fall into formal right uh fear of missing out you don't
23:03
right uh fear of missing out you don't
23:03
right uh fear of missing out you don't need AI if you don't identify the market
23:08
need AI if you don't identify the market
23:08
need AI if you don't identify the market and the real need for your gni feature
23:12
and the real need for your gni feature
23:12
and the real need for your gni feature okay and of course you have to
23:15
okay and of course you have to
23:15
okay and of course you have to understand and apply responsible AI
23:17
understand and apply responsible AI
23:17
understand and apply responsible AI principles and ethical implications of
23:21
principles and ethical implications of
23:21
principles and ethical implications of AI um such as um you have to understand
23:27
AI um such as um you have to understand
23:27
AI um such as um you have to understand your users you have to understand the
23:29
your users you have to understand the
23:30
your users you have to understand the data that you are sending out to people
23:33
data that you are sending out to people
23:33
data that you are sending out to people okay so here's uh uh this phrase from
23:37
okay so here's uh uh this phrase from
23:37
okay so here's uh uh this phrase from Robert Green the author of the 48 Laws
23:41
Robert Green the author of the 48 Laws
23:41
Robert Green the author of the 48 Laws of Power which is a fascinating book the
23:44
of Power which is a fascinating book the
23:44
of Power which is a fascinating book the best way to protect yourself is to be as
23:46
best way to protect yourself is to be as
23:46
best way to protect yourself is to be as fluid and formless as water never read
23:50
fluid and formless as water never read
23:50
fluid and formless as water never read on stability or lasting order Everything
23:54
on stability or lasting order Everything
23:54
on stability or lasting order Everything Changes okay so you have to adapt
23:58
Changes okay so you have to adapt
23:58
Changes okay so you have to adapt because this is a moving Target friends
24:01
because this is a moving Target friends
24:01
because this is a moving Target friends AI is a moving Target each day you can
24:05
AI is a moving Target each day you can
24:05
AI is a moving Target each day you can see in the news in LinkedIn and social
24:07
see in the news in LinkedIn and social
24:07
see in the news in LinkedIn and social media there are like thousands and
24:10
media there are like thousands and
24:10
media there are like thousands and thousands of different applications and
24:12
thousands of different applications and
24:12
thousands of different applications and Frameworks and options and
24:15
Frameworks and options and
24:15
Frameworks and options and models uh related to gen AI so this is a
24:18
models uh related to gen AI so this is a
24:18
models uh related to gen AI so this is a moving
24:19
moving
24:20
moving Target Technologies and techniques is
24:22
Target Technologies and techniques is
24:22
Target Technologies and techniques is speaking of those you have to understand
24:24
speaking of those you have to understand
24:25
speaking of those you have to understand models you have to understand the the
24:27
models you have to understand the the
24:27
models you have to understand the the basics of uh and essential traits of
24:31
basics of uh and essential traits of
24:31
basics of uh and essential traits of models and of course Frameworks that you
24:34
models and of course Frameworks that you
24:34
models and of course Frameworks that you want to use in your applications for
24:36
want to use in your applications for
24:36
want to use in your applications for building this kind of solutions rack
24:39
building this kind of solutions rack
24:39
building this kind of solutions rack Vector databases and
24:41
Vector databases and
24:41
Vector databases and glossery so let's uh talk about models
24:46
glossery so let's uh talk about models
24:46
glossery so let's uh talk about models I'm talking about large language models
24:48
I'm talking about large language models
24:48
I'm talking about large language models and also small large I'm sorry small
24:52
and also small large I'm sorry small
24:52
and also small large I'm sorry small language models I'm always confused
24:54
language models I'm always confused
24:54
language models I'm always confused about that L small language models
24:58
about that L small language models
24:58
about that L small language models so for instance if you want to implement
25:02
so for instance if you want to implement
25:02
so for instance if you want to implement this GNA feature in your application you
25:04
this GNA feature in your application you
25:04
this GNA feature in your application you have to
25:05
have to
25:05
have to identify if this super huge model is
25:11
identify if this super huge model is
25:11
identify if this super huge model is going to be enough or maybe this other
25:14
going to be enough or maybe this other
25:14
going to be enough or maybe this other smaller mod uh model is is better right
25:18
smaller mod uh model is is better right
25:18
smaller mod uh model is is better right um and you can also um identify and you
25:21
um and you can also um identify and you
25:21
um and you can also um identify and you need to identify closed source and open
25:24
need to identify closed source and open
25:24
need to identify closed source and open source models for example open AI High
25:28
source models for example open AI High
25:28
source models for example open AI High models are close Source right such as
25:32
models are close Source right such as
25:32
models are close Source right such as gpt3 and GPT 4 those are uh closed
25:37
gpt3 and GPT 4 those are uh closed
25:37
gpt3 and GPT 4 those are uh closed source and you can also use open source
25:40
source and you can also use open source
25:40
source and you can also use open source models such as Lama or mistol or some
25:44
models such as Lama or mistol or some
25:44
models such as Lama or mistol or some others uh that are as I said open source
25:49
others uh that are as I said open source
25:49
others uh that are as I said open source and you also need to identify if a
25:51
and you also need to identify if a
25:51
and you also need to identify if a general purpose or a specialized model
25:54
general purpose or a specialized model
25:54
general purpose or a specialized model is enough for you okay there are some
25:57
is enough for you okay there are some
25:57
is enough for you okay there are some models are General purpose and there
25:59
models are General purpose and there
25:59
models are General purpose and there there are some others that are
26:01
there are some others that are
26:01
there are some others that are specialized in a
26:03
specialized in a
26:03
specialized in a particular um industry such as healthare
26:07
particular um industry such as healthare
26:07
particular um industry such as healthare or marketing and so on so you have to
26:10
or marketing and so on so you have to
26:10
or marketing and so on so you have to identify that and finally you have
26:14
identify that and finally you have
26:14
identify that and finally you have to uh specify if want to use a software
26:18
to uh specify if want to use a software
26:18
to uh specify if want to use a software as a service solution such as open AIS
26:21
as a service solution such as open AIS
26:21
as a service solution such as open AIS end points or Asher or AWS and so on or
26:27
end points or Asher or AWS and so on or
26:27
end points or Asher or AWS and so on or maybe you want to host your own models
26:30
maybe you want to host your own models
26:30
maybe you want to host your own models in your own infrastructure or private
26:33
in your own infrastructure or private
26:33
in your own infrastructure or private Cloud
26:35
Cloud
26:35
Cloud right so I think that's better when you
26:38
right so I think that's better when you
26:38
right so I think that's better when you are trying to lower costs and you want
26:41
are trying to lower costs and you want
26:41
are trying to lower costs and you want to uh keep your data with you right and
26:45
to uh keep your data with you right and
26:45
to uh keep your data with you right and not let not let the data go out to this
26:48
not let not let the data go out to this
26:48
not let not let the data go out to this other uh external thirdparty uh
26:52
other uh external thirdparty uh
26:52
other uh external thirdparty uh company so just be aware that models
26:56
company so just be aware that models
26:56
company so just be aware that models come and go each day we're watching the
26:59
come and go each day we're watching the
26:59
come and go each day we're watching the news and social media and models are
27:02
news and social media and models are
27:02
news and social media and models are everywhere right thousands and thousands
27:05
everywhere right thousands and thousands
27:05
everywhere right thousands and thousands of
27:06
of
27:06
of models and here's an interesting chart
27:09
models and here's an interesting chart
27:09
models and here's an interesting chart that I found the other day in
27:11
that I found the other day in
27:11
that I found the other day in LinkedIn and on average 60 days down to
27:16
LinkedIn and on average 60 days down to
27:16
LinkedIn and on average 60 days down to 20 days in the last six months are the
27:19
20 days in the last six months are the
27:19
20 days in the last six months are the average that model is in the top uh in
27:22
average that model is in the top uh in
27:22
average that model is in the top uh in the in the charts right so for instance
27:26
the in the charts right so for instance
27:26
the in the charts right so for instance uh Gemini 1. five Pro blah blah blah it
27:31
uh Gemini 1. five Pro blah blah blah it
27:31
uh Gemini 1. five Pro blah blah blah it was uh just three days at number one
27:35
was uh just three days at number one
27:35
was uh just three days at number one just three days in the charts so we're
27:39
just three days in the charts so we're
27:39
just three days in the charts so we're not talking about the you know music
27:41
not talking about the you know music
27:41
not talking about the you know music charts we're not talking about the
27:44
charts we're not talking about the
27:44
charts we're not talking about the billboard where some albums stay in the
27:48
billboard where some albums stay in the
27:48
billboard where some albums stay in the charts for years and years no this is
27:52
charts for years and years no this is
27:52
charts for years and years no this is quite different right uh just take a
27:55
quite different right uh just take a
27:55
quite different right uh just take a look at this one uh number of days at
27:59
look at this one uh number of days at
27:59
look at this one uh number of days at number one just one and of course this
28:01
number one just one and of course this
28:01
number one just one and of course this is because this is a new one chat
28:04
is because this is a new one chat
28:04
is because this is a new one chat gp40 this is U another one has that has
28:08
gp40 this is U another one has that has
28:08
gp40 this is U another one has that has stayed for a long time in the charts you
28:11
stayed for a long time in the charts you
28:11
stayed for a long time in the charts you get the idea just be aware of models
28:15
get the idea just be aware of models
28:15
get the idea just be aware of models they come and go you have to identify
28:17
they come and go you have to identify
28:17
they come and go you have to identify and you have to test those models just
28:20
and you have to test those models just
28:20
and you have to test those models just to see if they could work for your
28:24
to see if they could work for your
28:24
to see if they could work for your Solutions talking about
28:27
Solutions talking about
28:27
Solutions talking about Frameworks so there are many many dozens
28:30
Frameworks so there are many many dozens
28:30
Frameworks so there are many many dozens of different Frameworks out there okay
28:33
of different Frameworks out there okay
28:33
of different Frameworks out there okay so I'm just uh including four here
28:37
so I'm just uh including four here
28:37
so I'm just uh including four here semantic kernel which is a fantastic
28:40
semantic kernel which is a fantastic
28:40
semantic kernel which is a fantastic framework for creating um um llm based
28:45
framework for creating um um llm based
28:45
framework for creating um um llm based solutions by using net and python you
28:50
solutions by using net and python you
28:50
solutions by using net and python you can also use hyack or llama index or
28:53
can also use hyack or llama index or
28:53
can also use hyack or llama index or Lang chain uh for building those
28:56
Lang chain uh for building those
28:56
Lang chain uh for building those Solutions what's cool about semantic
28:58
Solutions what's cool about semantic
28:58
Solutions what's cool about semantic kernel is that if you are a c developer
29:01
kernel is that if you are a c developer
29:01
kernel is that if you are a c developer uh semantic kernel is a first class
29:05
uh semantic kernel is a first class
29:05
uh semantic kernel is a first class class citizen in the net world and Lama
29:09
class citizen in the net world and Lama
29:09
class citizen in the net world and Lama index Lang chain and H tack you can use
29:11
index Lang chain and H tack you can use
29:11
index Lang chain and H tack you can use Python and you can also use other
29:14
Python and you can also use other
29:14
Python and you can also use other programming languages rag retrieval
29:17
programming languages rag retrieval
29:17
programming languages rag retrieval augmented generation this is a very
29:19
augmented generation this is a very
29:19
augmented generation this is a very famous pattern and technique for using
29:21
famous pattern and technique for using
29:21
famous pattern and technique for using your own data to uh give context to the
29:27
your own data to uh give context to the
29:27
your own data to uh give context to the intelligence
29:28
intelligence
29:28
intelligence so it can respond back with uh drowned
29:31
so it can respond back with uh drowned
29:31
so it can respond back with uh drowned data in other words it can respond back
29:33
data in other words it can respond back
29:33
data in other words it can respond back with uh the data and information that
29:37
with uh the data and information that
29:37
with uh the data and information that you want your private data and documents
29:41
you want your private data and documents
29:42
you want your private data and documents so this is another pattern and this is
29:44
so this is another pattern and this is
29:44
so this is another pattern and this is another uh thing that you have to um be
29:48
another uh thing that you have to um be
29:48
another uh thing that you have to um be on the lookout H because there are many
29:50
on the lookout H because there are many
29:50
on the lookout H because there are many different options and different
29:52
different options and different
29:52
different options and different techniques for doing this and of course
29:55
techniques for doing this and of course
29:55
techniques for doing this and of course Vector databases Vector database say
29:58
Vector databases Vector database say
29:58
Vector databases Vector database say allow you to store embeddings which is a
30:01
allow you to store embeddings which is a
30:01
allow you to store embeddings which is a mathematical representation of words and
30:05
mathematical representation of words and
30:05
mathematical representation of words and the the explanation is more detailed
30:08
the the explanation is more detailed
30:08
the the explanation is more detailed than that but it actually a vector
30:10
than that but it actually a vector
30:10
than that but it actually a vector database allows you to store uh a vector
30:14
database allows you to store uh a vector
30:14
database allows you to store uh a vector which is uh you know in the cartisian
30:16
which is uh you know in the cartisian
30:16
which is uh you know in the cartisian plan U chart um you you can sore those
30:20
plan U chart um you you can sore those
30:20
plan U chart um you you can sore those big numbers and you can retrieve those
30:23
big numbers and you can retrieve those
30:23
big numbers and you can retrieve those uh
30:25
uh
30:25
uh those data and information from the
30:27
those data and information from the
30:27
those data and information from the dictor Vector databases in many
30:29
dictor Vector databases in many
30:29
dictor Vector databases in many different ways uh such as uh you can use
30:34
different ways uh such as uh you can use
30:34
different ways uh such as uh you can use aure AI search chroma pine cone many
30:36
aure AI search chroma pine cone many
30:36
aure AI search chroma pine cone many different options that we have uh
30:38
different options that we have uh
30:38
different options that we have uh nowadays and finally you have to be
30:41
nowadays and finally you have to be
30:41
nowadays and finally you have to be careful and you have to be aware of uh
30:45
careful and you have to be aware of uh
30:45
careful and you have to be aware of uh hundreds of different uh words and uh
30:49
hundreds of different uh words and uh
30:49
hundreds of different uh words and uh terms and uh little phrases about uh gen
30:54
terms and uh little phrases about uh gen
30:54
terms and uh little phrases about uh gen AI such as fine tuning zero shot Fusion
30:58
AI such as fine tuning zero shot Fusion
30:58
AI such as fine tuning zero shot Fusion and so on so I totally recommend to you
31:02
and so on so I totally recommend to you
31:02
and so on so I totally recommend to you you understand the basic right you
31:06
you understand the basic right you
31:06
you understand the basic right you understand the the essential glossery of
31:10
understand the the essential glossery of
31:10
understand the the essential glossery of gni if you want to create this kind of
31:13
gni if you want to create this kind of
31:13
gni if you want to create this kind of solutions you don't have to be an expert
31:15
solutions you don't have to be an expert
31:15
solutions you don't have to be an expert in the Transformer
31:17
in the Transformer
31:17
in the Transformer architecture um and actually I truly
31:20
architecture um and actually I truly
31:20
architecture um and actually I truly believe you don't have to be an expert
31:22
believe you don't have to be an expert
31:22
believe you don't have to be an expert in deep learning and you don't have to
31:25
in deep learning and you don't have to
31:25
in deep learning and you don't have to be an expert in machine learning as well
31:29
be an expert in machine learning as well
31:29
be an expert in machine learning as well but actually actually there's a nice
31:32
but actually actually there's a nice
31:32
but actually actually there's a nice example uh this is a web page that you
31:34
example uh this is a web page that you
31:35
example uh this is a web page that you can navigate to to understand how
31:37
can navigate to to understand how
31:37
can navigate to to understand how Transformers work and Transformer is an
31:41
Transformers work and Transformer is an
31:41
Transformers work and Transformer is an architecture that is used by the GPT
31:44
architecture that is used by the GPT
31:44
architecture that is used by the GPT kind of uh models and other um models as
31:49
kind of uh models and other um models as
31:49
kind of uh models and other um models as well so this is a pretty pretty cool web
31:52
well so this is a pretty pretty cool web
31:52
well so this is a pretty pretty cool web page that you can use to understand the
31:54
page that you can use to understand the
31:54
page that you can use to understand the the math behind a Transformer
31:58
the math behind a Transformer
31:58
the math behind a Transformer okay so use cases use cases the top four
32:03
okay so use cases use cases the top four
32:03
okay so use cases use cases the top four use cases that mck McKenzie and Company
32:07
use cases that mck McKenzie and Company
32:07
use cases that mck McKenzie and Company detected some months ago are software
32:12
detected some months ago are software
32:12
detected some months ago are software engineering customer
32:14
engineering customer
32:14
engineering customer operations product research and
32:16
operations product research and
32:16
operations product research and development and marketing so those four
32:20
development and marketing so those four
32:21
development and marketing so those four are the top um use cases where you can
32:24
are the top um use cases where you can
32:24
are the top um use cases where you can use gen
32:26
use gen
32:26
use gen AI uh so you can see there are of course
32:30
AI uh so you can see there are of course
32:30
AI uh so you can see there are of course other use cases but those four are the
32:33
other use cases but those four are the
32:33
other use cases but those four are the most important ones
32:37
most important ones
32:37
most important ones challenges so don't trust the models uh
32:42
challenges so don't trust the models uh
32:42
challenges so don't trust the models uh blindly okay they
32:45
blindly okay they
32:45
blindly okay they hallucinate they're trying to be gentle
32:47
hallucinate they're trying to be gentle
32:47
hallucinate they're trying to be gentle they're trying to be helpful so there
32:51
they're trying to be helpful so there
32:51
they're trying to be helpful so there are times where they just make up things
32:55
are times where they just make up things
32:55
are times where they just make up things right they hallucinate that's the term
32:58
right they hallucinate that's the term
32:59
right they hallucinate that's the term so that's a challenge culture you have
33:02
so that's a challenge culture you have
33:02
so that's a challenge culture you have to have the culture you have to have the
33:04
to have the culture you have to have the
33:04
to have the culture you have to have the team topology to create this kind of
33:07
team topology to create this kind of
33:07
team topology to create this kind of solutions maybe maybe you have to change
33:10
solutions maybe maybe you have to change
33:10
solutions maybe maybe you have to change the way you work nowadays if you want to
33:14
the way you work nowadays if you want to
33:14
the way you work nowadays if you want to adopt and you want to uh Implement jna
33:18
adopt and you want to uh Implement jna
33:18
adopt and you want to uh Implement jna in your software
33:20
in your software
33:20
in your software Solutions so you have to you have to
33:23
Solutions so you have to you have to
33:23
Solutions so you have to you have to have the the Buy in from the you know
33:26
have the the Buy in from the you know
33:26
have the the Buy in from the you know the the the compy itself and the
33:29
the the the compy itself and the
33:29
the the the compy itself and the stakeholders and maybe the culture um
33:34
stakeholders and maybe the culture um
33:34
stakeholders and maybe the culture um will be uh will need to be changed um
33:39
will be uh will need to be changed um
33:39
will be uh will need to be changed um and
33:40
and
33:40
and finally don't let analysis paralysis to
33:44
finally don't let analysis paralysis to
33:44
finally don't let analysis paralysis to stop you from creating something right
33:48
stop you from creating something right
33:48
stop you from creating something right begin now start now there are many many
33:53
begin now start now there are many many
33:53
begin now start now there are many many Technologies and platforms and models
33:55
Technologies and platforms and models
33:55
Technologies and platforms and models and
33:56
and
33:56
and Frameworks out there that you can use
33:59
Frameworks out there that you can use
33:59
Frameworks out there that you can use today okay don't let analysis paralysis
34:03
today okay don't let analysis paralysis
34:03
today okay don't let analysis paralysis to stop you
34:06
to stop you
34:06
to stop you and it was so so so great to have to to
34:09
and it was so so so great to have to to
34:10
and it was so so so great to have to to be here with you friends and enjoy the
34:14
be here with you friends and enjoy the
34:14
be here with you friends and enjoy the rest of the conference
34:16
rest of the conference
34:16
rest of the conference [Music]