We talk about understanding your business problem before building AI solutions, turning to the open source community for help, and building teams in the lens of the data science triangle.

0:00

Why are you looking into AI?

Are you formal driven?

Are you hype driven?

Did your, you know, stakeholder tell you like we should have an experiment in AI?

And I hear this a lot, It's these are like, you know, things that happen in practice or, you know, do you have a problem that needs solving?

0:22

Fractional is the show for the best consultants in AI design, software engineering, and product.Learn from the mental models of top fractional makers and founders.In this episode of Fractional, we talk to Tim Santos.

0:39

Tim is the Director of Product for AI Cloud Solutions at Graphcorp where he leads to product strategy and vision for enabling developers plus AI and MO software organizations to build and deploy cutting edge AI models in production.

0:56

Tim also leads the AI Hive on Swarm.With over 13 years in experience in Data Science, AI and ML OPS, Tim has a strong background and passion for delivering impactful and scalable AI products and services to solve real world problems and create value for customers and stakeholders.

1:18

In this episode of Fractional, you talk about understanding your business problem before building AI solutions, training to the open source community for help, and building teams in the lens of the Data Science Triangle, this is your host, Alexis Goliado from Swarm.

1:35

Enjoy the episode Tim.Welcome to the Fractional Podcast.Thanks Alexis.Thanks for having me.Can you please introduce yourself to everyone listening in?Well, hi, I'm Tim.I'm Director of Product from Graph Core.

1:52

I look after AI cloud solutions.So what I do is, you know, we have a tech ecosystem for AI, and I help out our community, our customers, to build solutions, build their startups on our ecosystem, Yeah, And I'm currently the hive leader of the AI hive in swarm, the AI community.

2:20

And I've been in in the AI space, machine learning, data science in like 13 years.Before it was cool.Before it was.Cool.Yeah.Just just you know, you should be like bunch of nerds, you and your mates know about it.

2:38

But now it's really exciting times because everybody talks about it.So there's some momentum in the in terms of being able to solve problems with AI.OK.So let's jump right in.So the first question for you, Tim, the Hive leader for AI.

2:56

Let's assume you're talking to a founder and they want to integrate AI into into their startup.Or even if you're a corporate innovator, right, like you have an existing business, like what considerations should you have to use?

3:16

AI Yeah, interesting question.So I think first I guess you should be asking yourself why?Why?Why are you looking into AI, right?You know, are you formal driven?

3:34

Are you hype driven?Did your, you know, stakeholder tell you like we should have an experiment in here?And I hear this a lot, it's these are like, you know, things that happen in practice or, you know, do you have a problem that needs solving or do you have a particular expertise you know you really need to build or want to build something in the AI space?

4:00

If you're in a more deep Turk researching space, once you've asked yourself those you have particular answer to it, then you could move on to the next steps.What are the next?Steps and I think we could focus on.

4:16

Let's say you have you have identified a problem and you have been aware.Let's say that you know there's this kind of pain point, There's an AI solution, or this is a typical problem that AI can solve.

4:34

OK, but that's warm.We value being problem centric.Yeah, right.And how does a founder or a corporate innovator start with that?Like how do you start with the problem?

4:50

Yeah, yeah.OK.So you are aware of AI you're trying to as a start up, right?Like you're trying to solve a particular problem.That's why you got into this whole, you know, space.What I've seen as as one way to sort of find potential use cases for AI is you know you you map out maybe your business processes, you find inefficiencies here and there just like any other technology you could be using this technology to either optimize your cost, you know optimize or solve inefficiencies or create value.

5:30

Usually when when I talk to to builder CTOS or or corporate so I I'd walk through some of their business processes and then find pain points or problems to be solved from there and then once you've identified a problem then you can start ideating what are the potential solutions.

5:58

But why would you want to take this approach versus just, you know, following the hype, right?Like chat bots or like using a GPT And like most people will go that direction, right?Yeah, you're going to follow the hype.Yeah, yeah.

6:13

What's?The trade off?You cannot blame people because that's, I mean that's the exciting tech that comes out in the media.That's what everybody's talking about and it's it's really cool like but as a business as you know you're trying to provide value through a solution for a technology or an AI solution then stay as you said like Swarm is product focused.

6:41

You know we have the saying in the product space wherein I'll fall in love with the problem because that that becomes your anchor and then you know you could find the appropriate AI, AI solution to to sort of solve that.

7:03

One of the the benefits of like the recent advances in AI is that these large language models, foundational models, they're they're considered as a Swiss army knife of machine learning tool right, that it can solve a variety of problems.

7:25

Actually it's it's much easier to sort of start developing or finding the right tools when you have that that clear target your clear goal.Because otherwise, and I've seen this a few times where in I know an executive would say like let's come up with an AI experiment, where do I go from here, like what's that, what's that like.

7:49

So I think I think that's a good kicker.I think that that pushes people to start, OK we've got buy in from up top let's find a problem then we start searching for potential solutions.Sometimes you know you could find incremental value from from places that that you don't usually expect.

8:15

So like I've seen some of the interesting or at least AAI solutions that have been deployed to like millions of of of customers being like something simple like you know, data quality right, Like it's it's not as sexy as you know.

8:33

Spell checker.Glorified spell checker or something like that.But you know, especially if you're starting to build success stories to get ROI from that time and and investment of building AI, those small wins go a long way.

8:52

You know you derive value from from this tiny thing, and then that becomes your.It's.Like your momentum builder.Yeah, yeah.Your brownie points to the next.Yeah, exactly.To get into more exciting stuff, cuz I mean takes.

9:10

There's some level of complexity in building AI.So the quicker you get value or you know the quicker or smaller the experiments are, the more iterations you could get.And you know, the faster you get to building cool and interesting stuff.

9:26

Summary you want to map out your business first.What's the current process?Where are the pain points?And like after taking a look, you want to identify how can AI step in and either reduce costs or streamline efficiencies or create value and drive revenue.

9:48

Yeah, yeah.If if you know the answer and you have the, you know AI capability, that's great.But if you're new in the space then you you have problem and then you approach do research and all of the there's there's there's a website called there's an AI for that because that's people start there or you know ask the community.

10:13

A lot of AI builders in the space has been empowered and and like put in the spotlight and are willing to share you know the stuff that they've been building for for a few years right.So yeah this is those.Use those resources.

10:28

The open source community like particularly in AI is is is very there.There's a lot of movement and and and collaboration and contribution.AI is open open source space.

10:45

Define your problem.Fall in love with your problem, then start looking for help doing research and then you can you can start, I don't know, building.You've done this for a while, for a lot of different companies and different verticals as well, and you're obviously stressing the value of seeking help from community, right?

11:09

Can you tell me more about that?Like why should you know founders or corporate innovators?Like, talk to, you know, community leaders like you.It's a very nascent space, a lot of changes happening, a lot of new things being developed, innovation, I would think that you know, there's also also, you know, a lot of noise that comes in because of that.

11:32

So AI is a complex, you know, technology on its own, your business, your, your problem space is also complex on its own.So being able to map out the problem with the solution space and those who have had that experience could could point out especially deploying into the different domains in verticals, they could point out that well actually in your space this AI component could solve that like.

12:00

That's an example.That's an example if we are in the healthcare space, right?Let's say diagnostics, right?Like X-ray computer vision that detects a particular disease, that could be one.But there are the components in in in the healthcare space like NLP for patient care.

12:22

So if someone's, you know, describing the ailments and all that that's a generative AI use case you have.You know that could be that could fall into like that OK, chat bot kind of solution.But there are other parts of that space as well, like, you know, a doctor's prescription, you have an OCR solution that can sort of make it easier for people to read the doctor's presumptions, right?

12:50

You know, if you're just going to stay in the sort of this hype space, then you wouldn't have identified other parts of the potential value that AI can bring in.So.

13:06

And that's where strategists.Yeah, Yeah.That's where the strategies come in or you know advisors that can sort of translate right translate the business problem to the solution space, the AI solution space as well would be patient enough to like go through you know finding out the, the path to to to to success in adopting a particular technology so.

13:36

Let's say you you've got the use cases down, you've you have a strategy, you have a general direction on where you want to go.Can you talk to me about, you know, just building teams, like how complex is that?Like how hard is it as a as a as a problem?

13:52

This data science triangle that people in the data science space use a lot wherein to to deliver data-driven data science and AI solution is to have one the the domain expertise the main subject matter for the healthcare.

14:16

Yeah for example your healthcare practitioner all the insurance and outs and I was in in finance.So if you're a tax person they can tell you that well actually this is what tax problems are how we do it.How the process works.

14:32

You you need the other pillar to that sort of triangle is the the algorithmic knowledge mathematics you know yeah being able to to make sense of or you know codify the the problems into to algorithms and and the software could to complete that sort of triangle as a software person.

14:59

So the maths with the software allows you to put something into a production let's say, or make it at least production ready rather than just an experiment or I guess.

15:16

This is where hiring consultancy teams come in because it's really made like you have a software guy, a machine learning or maths person and I guess a founder or a client would serve as the main expert.

15:33

The challenge is that it's three different skills.Ideally, if there's one person that can do all three of them, you know you're the main expert that can code and knows AI, machine learning and science, right?

15:52

Algorithms, yeah, yeah.So those are very sought after resources and they will be expensive, right.But at the same time I think they would still have 24 hours in a day or 88 hours in no work day.

16:12

So there will still be limit to to what what they can do.And then it's that's why this, this data science triangle is is like one of the building blocks when when it comes to either building your capability or building an AI product or a data science project.

16:32

So yeah.Yeah, that's an awesome mental model to have.Like the data science triangle.Yeah.One last thing I want to ask you as we end this conversation, What's your general advice to all the founders and corporate innovators who want to get into AI?

16:50

It's exciting time.So now it could get overwhelming, but but the space is very, you know, very nice.And so a lot of things are changing.A lot of, you know, like the LLM leaderboards in the last six months, you know, one after another every couple of weeks there's a new thing that comes along.

17:11

But there's there's also value to be there's, there's a reason for that excitement and and hype, right.So there's value to be, to be made.Fall in love with a problem.Start with that, And then there's a lot like too many solutions that are available out there sometimes.

17:29

Yeah, yeah, and complex systems require more than the collaboration from multiple skilled individuals, right?So don't be afraid to to use the resources of the community, the the open source, you know tooling models and all that experimenting and then you can start experimenting and and finding the right value from from AI.

18:02

So yeah, I think it's more or less my.Awesome.So that's Tim, our AI Hive leader at Swarm.And thanks for coming on to Fractional.Yeah.Thank you.

18:17

Thanks, Alex.