Dr. Eva Agapaki, an AI product leader, discusses the current state and future of AI, as well as the challenges and opportunities for women in the field. She emphasises the importance of generating new information and context through generative AI and highlights the need for AI governance and regulation.

Dr Agapaki also shares her journey into AI and the inspiration behind starting Hatch Labs, a company that helps entrepreneurs and organisations adopt AI. She also discusses the misconceptions about AI and the potential for AI to augment human work.

Takeaways

• AI governance and regulation are crucial to address ethical concerns, biases, and potential misuse of AI.
• Women in AI face challenges but can find support through mentorship, networking, and conferences.
• AI products and pilots require careful consideration of scalability and integration with existing systems.
• The future of AI holds opportunities for innovation and positive impact but also requires responsible and ethical use.
• Aspiring AI professionals and entrepreneurs should attend conferences, join communities, and stay informed through newsletters and podcasts.

00:04 - Joanna Shilton (Host)

Hello and welcome to Women WithAI, the podcast dedicated to amplifying the voices and perspectives of women in the ever-expanding field of artificial intelligence. Today, I have the pleasure of speaking with someone with over eight years of experience at the forefront of AI and machine learning innovation. She has a background in civil engineering, but her deep technical expertise is rooted in a PhD in computer science from the University of Cambridge and MIT, complemented by her role as an assistant professor at the University of Florida. But before we get started with the podcast, let me tell you a little bit more about her.

00:36

Dr Eva Agapaki is an AI product leader adept at building B2B and B2C product portfolios that cater to a spectrum of startups across many different industries. Her eight years of experience at the forefront of AI and machine learning has emphasised a customer-first, data-driven approach. She is the founder and CEO of Hatch Labs, where she advises deep tech founders and organisations on building AI-powered products. Her biggest achievements to date including launching million-dollar revenue-generating products and establishing innovation labs in startups and large organisations from conception to market in highly regulated industries such as aviation, oil and gas and ad tech. While teaching at the University of Florida, Eva built a 1.5 million dollar research portfolio, validating early-stage breakthrough ideas and partnering with cross-functional teams. She's received numerous AI leadership awards and is a featured speaker at many AI conferences. Dr Eva Agapaki, welcome to Women with AI.

Hi, Jo, thank you for hosting me.

Oh, it's brilliant. It's great to have you here, and can I just say wow, so you're at the.

01:42

University of Cambridge over here in the UK, and then MIT, and you've taught at the University of Florida. So, as an expert for anyone who might be new to the podcast, please can you explain to us what is AI?

01:56 - Eva Agapaki (Guest)

Sure, yeah, that's a word that actually we hear a lot every day and, just to give some context to our audience, when we refer these days to AI, we usually refer to generative AI or gen AI, so particularly like how we can generate new information or context from a lot of different inputs, so be it text, audio, video. So that's the main differentiating factor, as opposed to everything before that, which was like ML and deep learning, where it was everything around predictive analytics. It was everything around predictive analytics, and now we have more depth in synthesising information, so that's what we refer to.

02:54

03:41

That was the subject of my PhD PhD thesis, which I then commercialised, and actually, that was the first introduction that I had into startups and being an entrepreneur, which I really loved, and starting building product teams zero to one.

03:58

That has been the main driver in my fashion over the years. And then I expanded that through different industries by starting my own research lab at the University of Florida. This was in partnership with NVIDIA, with a very large research grant, to how we can use AI in education, but also in different industries, so early prototypes in, for example, how we can maintain infrastructure, meaning like our bridges, highways, all up to like healthcare and oil and gas industries. So, I've always been like passionate about driving impact in industries where the application of AI is not yet achieved. So it takes like a while to get there, and all of these experiences actually inspired me to start Hatch. So Hatch Labs was conceived out of my inner passion to help other entrepreneurs and organisations that are on their path to identify how they can adopt AI and bridge that gap between research in Gen AI and application.

05:24 - Joanna Shilton (Host)

Are you seeing sort of trends in AI? Like do you know all the people that you work with? What's sort of the? Yeah, how do you stay ahead of the curve, as it were?

05:35 - Eva Agapaki (Guest)

Yes. So there is a ton of applications when it comes to generating content. So we will see like a ton of consumer-centric apps when it comes to like generating marketing campaigns or improving, like sales processes, these types of processes, and I use some of these as well, like when it comes to driving growth for hatch, um and uh and getting all my action items for meetings, um, content creation and all of that um. However, where I see, like, um, new, like trends, new trends are in reducing costs and especially in the domain of prototypes and content and use case validation for the industries that I mentioned earlier, like prop tech and aviation manufacturing. For the industries that I mentioned earlier, like prop-tech and aviation manufacturing, where it takes a lot of time, like in many cases, like five to 10 years, to actually see a product in the market and being used at scale, there is a whole process that we need to apply, and this is what I've been doing at Hatch and also what I'm seeing in these industries to be able to fast track that five to 10-year journey into a few couple of years.

07:46

I see two main trends. First of all, to reduce the costs by introducing small language models, like the Fee3 that we've seen. It could be any of the well-known model LLM providers like GPT or cloud and so forth, and actually execute them at scale. So there are startups, like Fiddler AI, for instance, that help with running these models in production, but I see that as an ongoing trend to improve product launches and being able to run these applications at scale. And the other trend that I have observed in different industries is when it comes to AI governance principles and actually regulating all the hallucinations that we get, the biases, and how we can continuously keep monitoring these systems without creating any type of fraudulent behaviours and keeping track of all the misunderstandings or misleading information that we may get.

08:52 - Joanna Shilton (Host)

Yeah, and again, because how do you, how would you sort of approach that? I mean, when you're building AI-powered products, or you're helping bring them to market, how, how is that regulated? Because that's some sort of, you know, the highly regulated industries already. So, yeah, who's regulating the AI? What? How does that?

09:11 - Eva Agapaki (Guest)

work, so there are a lot of ongoing efforts with regards to that, of course, country-level efforts like there are many, many countries that are in the process of creating these regulatory bodies, in conjunction with big tech companies and smaller AI startups as well. We are still in very early stages of that. When it comes to implementing these AI models and regulatory bodies, since these are new, as I'm calling them, research products, it takes a while to actually be able to regulate them. This is a one of the hot topics that we're going to see in the coming years.

10:20 - Joanna Shilton (Host)

uh, as they see it in discussions with different entities, I guess it's regulation, and that does that encompass, like, ethical concerns and bias and that kind of thing absolutely yes, uh, so we have already, like in the EU Act, when it comes to applying AI into different industries, regulatory requirements.

10:54 - Eva Agapaki (Guest)

So that will need to comply with the overarching governance of AI and what is allowable or not, and, of course, within the spectrum of these applications in the respective industries.

11:14 - Joanna Shilton (Host)

I'm talking about different industries and that kind of thing because we are here on women with AI. What's it like to be a woman working in this space?

11:24 - Eva Agapaki (Guest)

So, yeah, that's a great question since there are not so many women in AI, and, speaking from my own experiences, I've always felt like I'm the only woman in the room when it comes to these discussions and applications, so it's always interesting how to inspire more women to actually enter the space, and I would say that because I'm now also close to investments in AI startups, so the VC world also is behind when it comes to not only women entrepreneurs but also women investors.

12:16

So there is definitely a long way that we need to cross, but I would say there is some hope in this tunnel, and this is what I want to leave our speakers with when it comes to, specifically, women in this field to pursue their dreams and really hone in a niche that they have carved throughout their paths or that they're working on right now.

12:53

To actually pursue a degree in stem and then follow their passion um, because it, despite the challenges um that may come through their way, uh, it's possible to actually succeed and get support from other women, and this podcast is also a great way to um to find other women in the space and create like small support groups, uh to help each other uh or um uh within their own, like organisation they're, they can have their own uh allies that can support them, or mentors, like I still have from my first job, like my mentor who was a woman and still is a source of inspiration and we stay closely in touch with. So I highly encourage other young women to partner or find mentors that can inspire them and give them like positive energy to navigate the space, either that's in corporate world or in entrepreneurial, their entrepreneurial journey journey.

14:17 - Joanna Shilton (Host)

Yeah, I think that's great advice, thank you, even thank you for your kind words about the podcast as well, because you're right that that's my mission, that I'm learning about AI with every amazing woman that I speak to, and I I'm just, you know, learning more and more and spreading the word and trying to put more women in touch with each other, and I think the idea of mentors is a really good idea and I think that you're right supporting each other, because I think all too often women don't shout enough or, you know, put themselves forward enough.

14:38

It's quite easy to to sort of just sit there and think that you'll be recognized for what you're doing, and actually I think we all need to support each other, to rise up and and to make sure that we are involved, because you know all the things about bias and ethics and that sort of thing. It's really easy for us to to not be mentioned and not be there, so we definitely need our seat at the table to make sure that we're there, and I sort of a lot of the people that well, some of the people I've been speaking to are sort of saying it's not just about getting you know girls and women into STEM. It's about actually being able to retain them. So do you see the sort of challenge or the sort of the future of AI? Do you sort of think there are those opportunities for women? How can we sort of you know, jump on those? Do you think how will it impact on women specifically?

15:28 - Eva Agapaki (Guest)

Yeah, that's a great point because we actually see, like the trends.

15:35

I recently saw an article about the I think it was even close to like 40, 42 percent, I think, of women um in tech that actually uh leave their, their careers, which is pretty high.

15:50

So I think this is most important from even like attracting more women into STEM, but by actually retaining the talent, the women talent, in these industries. And this goes back to my earlier point of actually creating like allies and groups of women that are actually supportive of each other. There are a ton of conferences where this is a great way to like mingle and meet with other like-minded women like um uh women in product, which is a very large one for for for women product leaders and product managers. Women in tech, uh like a lot of different ones, depending on interests of the audience, and these are great ways to actually not only stay on top of the industry but also like creating those circles that will be very important. And, of course, while we're now navigating the different complexities and challenges in the workforce, which is important to always have an ally or someone to navigate that with, because it's very different to have that discussion with a woman as opposed to a male professional. Yeah.

17:19 - Joanna Shilton (Host)

I think we do see things differently sometimes, don't we? Yeah? So, I think it's definitely important to have that equal sort of spread of knowledge. What are some of the biggest misconceptions about AI that you've encountered along the way, and how do you address those?

17:40 - Eva Agapaki (Guest)

That's a great question because I've recently encountered a lot of actually, since this Gen AI technology is so new there is a ton of misconception as to what is possible versus what is feasible. So there are a ton of challenges and business problems yet to be solved, and I believe that we've just scratched the surface of what can be solved. Like I mentioned there, there is a ton of opportunity when it comes to customer support, to all these like improved tedious processes that can be improved, when it comes to content creation and generation or even information retrieval with, like the newest systems that we're seeing, like ragged systems, which make this possible because, since we have a ton of data, then it's possible to actually retrieve them efficiently and fast. But when it comes to like this more complex business or engineering type of problems and challenges, we're still pretty early on when it comes to especially enterprise level products and and and we're just starting to to see the potential and the feasibility of these products. So I believe that there's a long way to go.

19:13

We're still in the beginning and we will see a lot of changes, and that's something that I've been very, very passionate about not only changing, because we've seen a ton of change in technology, but we haven't still adapted on the business side in terms of like, adoption and success rates when it comes to successfully launching these products to market, because it's a process it won't be um within just one year like it's very similar with well, with what we saw with the surge of dot com, when there was like a whole shift in in businesses in the whole world.

19:56

Essentially and this will happen gradually I don't believe that we're going to be losing our jobs and a lot of people are saying that or that SAS is going to be, is not going to be anymore as a service. Actually, I heard that recently of service as a software instead of software as a service, which is like a very I would say like far-fetched view on things. But I'm a strong believer that actually, with proper regulation and regulatory bodies formed, all these misuses of AI, if I would say, will be prevented and we're going to actually you get only the benefits of it and being able to regulate the rest yeah.

21:10 - Joanna Shilton (Host)

So I think, yeah, I think you're right. It's AI is going to augment our work, not not put us all out of a job or overtake us. You still need that human element, or it's there to help us, isn't it? It's a tool, exactly, rather than, yeah, something to be scared of. And what are your thoughts on sort of AI products and pilots that are sort of? I mean, are you seeing them being used at scale, or do you, you know, when you're launching them, are people picking up sort of the proof of concept, or what's your experience with that?

21:49 - Eva Agapaki (Guest)

So it depends on the industry because sometimes, with a proof of concept, there might be that the data set that we've used or the specific use case that we've picked up in, let's say, a specific region or a specific population and so forth it might not be representative of the problem at scale. So, and to give you an example from my own experiences in my previous startup that I had, we were very close into a pilot. We completed the pilot into deploying a digital twin factory, but actually at scale this was not possible to happen because of the market and the systems not being updated to actually encounter that level of data that we needed. So we had to take a step back and rethink about the problem. So it's not always a sign that a successful pilot will mean a great product at scale. It takes a while and maybe a lot of different pivots depending on the systems that the customers use in respective industries, Since these systems, like.

23:18

Many times I've seen that customers do not have their systems with the updated, like, let's say, cloud, cloud providers right now. So it's it's a whole digital transformation on its own before actually being able to deploy these latest tools at scale. So it's going to be like a whole process like like. I have developed through Hatch, a framework that I'm helping companies navigate these uncharted waters with, like the incubation phase, where we're really focused on like getting out the proof of concept or MVP, depending on the stage that they're at Hatch, which is like launching. And lastly, grow, which might be like even two years down the line. Um, because it's and especially with llms, it's not always um clear if you haven't seen the output and the, the way that users will react and interact with the product, uh, how successful it's gonna it's gonna be so it's continuous. That's why I mentioned earlier about the model observability, where we actually need to monitor these systems even after launch.

24:40 - Joanna Shilton (Host)

Yeah, it feels like there's a lot of work to do still which is exciting.

24:44 - Eva Agapaki (Guest)

It is yeah, there's a lot of space for innovation for sure.

24:49 - Joanna Shilton (Host)

Yeah, what excites you most about the future of ai, do you think?

24:54 - Eva Agapaki (Guest)

um, so the possibility to actually uh, recreate now, like as we had with sass and the of com, like recreate the future, like I think that this is a rare point in time where it's possible to actually innovate for the next couple of years and really leave a mark in how things will be later on, assuming that we use AI in an ethical and responsible way. So I'm really excited to actually being able to see startups solve important challenges and problems, not for good.

25:47 - Joanna Shilton (Host)

Are you able to get involved in the ethical side of things and the regulation? Is that something that you or anyone working in AI can influence?

26:02 - Eva Agapaki (Guest)

So there are definitely I know in the States there are fellowships with the White House that there is a more direct involvement in these policies or working with some departments of state to work with the regulatory bodies in terms of forming these regulations, with the regulatory bodies in terms of forming these regulations. I'm not directly involved into these efforts maybe in the future, but because right now I'm most excited about the application of AI at scale for these industries that need it the most because of the challenges I've seen over over the years yeah, well, you mentioned those, the conferences and things um that are out there.

26:56 - Joanna Shilton (Host)

And, as an ai leader and educator, which you are, what advice would you give to aspiring ai professionals and entrepreneurs?

27:04 - Eva Agapaki (Guest)

yeah, so, um, there are a lot of interesting research conferences um that, um, I recommend uh, our audience um checks out, and these are the popular um ICML, iclear and CVPR. These are the ones that I'm personally attending or have been contributing to. And then also all the major and big tech providers have conferences around Gen AI. So this is also important to check out and be in the lookout for those, because they organize a ton. I'm sure, like in london as well as like in new york, I've seen like a lot of those and I'm going to many of them them as as well, because it's any of them as well, because it's a great way to to interact with other like-minded professionals and and see what's the latest in terms of applications. And, lastly, there are a lot of communities of entrepreneurs and AI builders.

28:23

I would say I'm part of like and um ai builders. I would say, um, I'm part of like and so, looking through like linkedin in terms of what's available, uh, there is a lot happening in new york. Uh, if we have audience that's around, I'm happy to share like lists and uh events. Uh, that I'm getting a ton. It's actually very hard to get to all of them and there's definitely a spike in momentum and momentum in this right now. And lastly, newsletters. I have started a newsletter where I've been educating and I can share the link as well with our audience in terms of simply explaining all these concepts from LLNs when it comes to what we can think about a successful product launch in the first days until it becomes scalable, what's right, like all these concepts, like explained from a product and business point of view, uh and uh podcasts, like the one that we are here today fantastic.

29:40 - Joanna Shilton (Host)

We'll definitely put the links in the in the show notes here to your yeah, to the newsletter and to linkedin so people can find you and, um, I guess, in the spirit of the podcast and and sort of learning, uh, from each other, is there anyone that you'd like to see on the podcast or what? What sort of, what sort of people would you like to learn from? If I could, if there's anyone that I could get on the show, who would you like to or who do you recommend I speak to next?

30:06 - Eva Agapaki (Guest)

uh. So, uh, the product leader, who's also an investor and her name is natalia burina uh, she has been um um at big tech companies and also she's an entrepreneur, so definitely a very influential woman and we've done some workshops together when it comes to like achieving product market fit for founders, so you can check that too and yeah. I definitely recommend her.

30:40 - Joanna Shilton (Host)

That's really good. Great, thank you, Dr Eva Agapaki. Thank you so much for coming on. Women with AI it's been absolutely fascinating to speak with you.

30:49 - Eva Agapaki (Guest)

It was a pleasure to talk to you, thank you.