024 From NVIDIA to Gen AI Data Privacy with Dr. Eiman Ebrahimi of Protopia.ai

Dr. Eiman Ebrahimi, CEO and co-founder of Protopia AI, joins Your AI Roadmap to discuss how his company is revolutionizing the safe use of private data in enterprises. After a decade at NVIDIA, he co-founded Protopia AI. Protopia focuses on enabling secure data utilization for machine learning and generative AI applications, ensuring sensitive information such as intellectual property and financial records remains protected. Dr. Ebrahimi highlights the challenges businesses face when balancing data privacy with AI innovation, particularly in industries like healthcare and finance. 

Podcast:

🗳️ Nominate Podcast Guests! https://yourairoadmap.com/podcastguest

Quotes

🛡️ "We sometimes make the analogy of, you know how there's a, sometimes you have a privacy layer on your laptop screen or on your phone screen. And, at the angle that you're looking at it, you can see it, but somebody looking at it in an angle they shouldn't be won't be able to tell what's going on on the screen.”

⚙️ “I spent 10 years at NVIDIA ... my background was not yet focused on AI, I come from a field that people refer to as computer architecture.”

Takeaways

🔐 Data privacy is a top priority for businesses using AI tools.

⚙️ Protopia AI creates privacy layers to protect sensitive information during machine learning processes.

🎯 Generative AI and language models have rapidly changed how businesses utilize private data.

🚀 Continuous learning and curiosity are key to thriving in AI and tech.

Bio:

Dr. Eiman Ebrahimi is a Co-Founder of Protopia AI, and serves as its Chief Executive Officer and President.

Prior to Protopia AI, Eiman was a Research Scientist at NVIDIA for a decade, where he led research to solve key challenges in accessing massive datasets for Al and optimizing large scale GPU systems for LLMs and recommender systems.

Connect with Dr. Eiman:
https://www.linkedin.com/in/eiman-ebrahimi-b6224a40/

Transcript

0:00

our main... mission is to enable the safe use of private information, especially at the enterprise level, where much of the data that enterprises work with is in fact private to them. Hi, my name is Joan Palmiter Bajorek. I'm on a mission to decrease fluffy hype and talk about the people actually building in AI. Anyone can build in AI, including you. Whether you're terrified or excited, there's been no better time than today to dive in. Now is the time to be curious and future -proof your career, and ultimately, your income. This podcast isn't about white dudes patting themselves on the back. This is about you and me. and all the paths into cool projects around the world. So what's next on Your AI Roadmap? Let's figure it out together. You ready? This is Your AI Roadmap, the podcast. Hey folks, this is Joan popping in to say a little intro about the episode. So in this episode, we get to talk about data at companies. And maybe you've worked at a company where it's pretty clear that there are pieces of data siloed that no one really knows how to use effectively, but there are nuggets of gold hiding in those data sets and really protecting internal data. I think we'll often think about like, maybe downloading data from hugging face or, can we protect the data from outside malicious players? But are we leveraging the data inside a company at the company? that's what this conversation is all about. Let's dive in. Hello, hello. could you introduce yourself please? Yes, yes, First of all, thanks for having me. My name is Eiman Ebrahimi. I am the CEO and co -founder at Protopia AI and excited to be here today with you. Cool, well grateful to have you. Okay, so what are you all doing over there, Protopia AI? So Protopia is a company that's focused on enabling the use of private information with generally machine learning, but more recently in the past year or two, obviously, as the field has gotten very focused on language models and generative AI, that's also where we have seen a lot of use cases pop up. But our main... mission is to enable the safe use of private information, especially at the enterprise level, where much of the data that enterprises work with is in fact private to them. And that data is often where the value lies for them as well in terms of what they can actually get from a new technology that comes about. And so, Being able to use that in a safe manner is a main part of the story. And that's what we try to unlock with the products and platform that we built. Awesome. That's really cool. Well, and I think data privacy, but leveraging the insights of data, data is so powerful. Can you give us a specific example of, so it's this business to business enablement platform. I think when you mentioned data privacy, just to set the context a little bit broader, it's really interesting that having been in this domain of talking about private data, often first thing that comes to mind are the things that are super important we've thought about as an industry for a long time. Things like personally identifiable information, health information, financial records, all things that are very important. But one of the, I guess, new aspects of generative AI and language models has been that there are a lot of use cases that enterprises think of where now the use of information is a lot broader in terms of what private information they are wanting to use. It's a lot, it goes beyond really personally identifiable information or health records. And it starts getting into things like company financial records or starts getting into even intellectual property, design blueprints, trade secrets. These are the sorts of things that are a lot harder to pinpoint as well, meaning, You can't really take a piece of code, for example, that somebody is writing and put a finger on here's where the IP is in this piece of code. The code itself embodies that IP. And so that is private data that often with something like, for example, a code completion tool can help provide some value. But using that private data becomes something that the enterprise needs to think hard about in terms of how do I use this code completion tool without potentially leaking the private data that I have in this piece of code, right? And so that sort of use cases is an example of, for instance, in the technology sector, where use cases around code completion are very effective, but what data gets used with them, and what users are able to use these tools for becomes a question for the business. And that's where we try to enable the safe use of that sort of private information. That's so cool. let me certainly read that back to you and see if I got it because I think other people who are not coders or I love a little R in Python in my life, So in people do a lot of auto -complete with like emails or text messages where it looks at it and says, we think the last word is gonna be Friday, right? So the AI is looking back, you know, natural language processing and the other words in context and saying, it guesses the next words, generating new text. What I'm hearing, I think, in a coding setting, when I'm typing a line of code, it begins to look at the previous code to guess what the next pieces of code are. But by looking at it, could unfortunately uncover things and objects I'm not interested in sharing. Is that a plain language? yeah, in plain language that that is the use case. But when we talk about things that we as you as the person that owns that code may not want to share, the question often is not whether or not you don't want to share that with. the LLM application, the language model application or the generative AI application. It's not that you don't necessarily trust the application. Even when you trust the application, at the end of the day, applications run on some machine somewhere. for... No good reason often that will compromise a machine just because the username and password of somebody was phished and happened to be on the dark web somewhere. data that is being sent to that machine can leak. Now your data is basically on a compromised system when you are just using it as a person interacting with that application, you're sending code that way for it to come up with. say, helping you code faster, and yet your information is leaking on that system. So that's an example of something that's not even about sharing with the provider. We trust our providers often, and there's a lot of very well -thought -through guarantees about what the providers are doing. But in the world of machines being compromised, that's a thing that just happens. So what we are trying to do as a company is build additional layers of data protection, essentially, on the data that is sent to those machines, Okay, that makes sense. These kind of buffers of it as a privacy layer where you essentially change what the data looks like before it goes to that application such that the application can still work, but it's almost for the application's eyes only. If somebody else were to look, it wouldn't mean anything to them. We sometimes make the analogy of, you know how there's a, sometimes you have a privacy layer on your laptop screen or on your phone screen. And, at the angle that you're looking at it, you can see it, but somebody looking at it in an angle they shouldn't be won't be able to tell what's going on on the screen. You can think of it in that manner where the application you're sending it to in that analogy will be you looking at that screen and the application will be able to see what you sent to it, but somebody else that shouldn't be looking at that wouldn't be able to tell what's going on. That makes sense. Well, and as I listen to you, honestly, I feel like my blood pressure is rising. There's like anxiety of this privacy world and, you know, what applications to be trusted, in your marketing or with your customers, like, do you speak about kind of fear it feels like big emotions will be bubbling up in this sector, like regularly. Yeah, I think there's generally a lot of big emotions that bubble up around any evolving sector And what's most important for those of us that are building essentially the picks and shovels that make it all happen is to create some level of... confidence in the fact that the different places where things can become problematic are being identified and solved for. one of the biggest challenges in technology expansions, like what we're seeing with generative AI is if we, as a collective, don't consider that there are new challenges that appear. And so if you look, especially in industries like healthcare and industries like finance, where there's potential for a lot of new applications to be built and efficiencies to be brought forward, and even really exciting new solutions to come to market, a lot of that has been blocked by much of this sort of figuring out what's going to happen with private data, right? Because those are industries where there are very, very important guardrails in place and being able to unblock data to get to these new classes of applications has been something that companies like ours have been working on. creation of value for businesses. that's so cool. You're talking about guardrails and like this this new innovation. I've been looking at kind of technological revolutions And one of them has been really about the advent of the car. And when we went from horse and buggy to automobile, roads needed to be built, right? New laws needed to be passed. This infrastructure. But you could see it visually. You could see when these cars came to town and the price plummeted, very similar to Tesla life. And then suddenly these roads need to be built, infrastructure laws, blah, blah, blah, blah, blah. In this advent, we're seeing this infrastructure that needs to be built, safeguards, et cetera. But we don't actually see it necessarily. It's in these computers. do these companies see this infrastructure enablement that you're working on? these are really good questions. over the past, say 20 years, let's call it. individuals caring about the privacy of their own data is one topic that is super important when it comes to consumer interaction. For example, Apple is an example of a company that builds a lot of consumer products where they take privacy very seriously and they try to provide as private of experience as they can to customers. Right? How, how Well, they do that is for their customers to decide. But when it comes to businesses, the challenge is that businesses are often a point of aggregation of a lot of data. that's talking about personal data privacy, right? But then the organization has its own data that's private to it as well and its own secrets that are about the business itself. And those are also not necessarily controlled by some regulation unless we're talking about a public company and laws and regulations around that. So a lot of private information becomes a lot more important at the business level. And to that extent, businesses are looking for solutions like this because that whole stop, don't use anything. I imagine was happening in those days where cars were becoming something that people want to use more effectively and all the, well, how do the roads get built with the municipalities and what are the laws that we need to have and who's going to register these vehicles and... I imagine a lot of those questions were happening there as well. It's an interesting time to be part of that ecosystem build. For sure, for sure. Well, and how you mentioned how old is the company? We've been around for four years now. Okay, nice. Can you tell us more about kind of as you've built the company Would love to hear about, surprises and learnings you've had along the way. Yeah, so when I say four years, it's interesting because when we started working on Protopia, generative AI and language models had not yet burst into the market the way that we are here today, right? That was maybe halfway into our journey as a company. So when we started building this notion of... let's build a platform that enables people to use private data with machine learning. We were actually just talking about machine learning and deep learning prior to late 2022, 2023. And so a lot of the use cases that we were involved with or our customers would think about were things, for example, in computer vision, where it would be, for example, a a system to be able to track the number of people coming in and out of a building in a smart city application or being able to look at a train platform and identify, for example, what the foot traffic coming and going into this particular particular part of the train platform is. Again, something computer vision wise. And so a lot of what we were seeing our customers imagine for how they would use these picks and shovels were around these sort of smart cities, smart manufacturing type of use cases. And then suddenly towards the end of 22, all of those same customers got very, very interested in language models and documents and text. And it was that that was a kind of very quick change in. the market and where we started focusing our energy on, even though our product, the software, the actual thing that people used to enable the use of those end models, that software is the same software as it was before, but then how we saw customers start wanting to use it suddenly changed. And it was in a matter of two months that we went from hearing a lot about computer vision to almost only 10 % people talking about computer vision and everyone being interested in how do I use this for interacting with a language model. So that's the sort of very quick dynamic change of what customers wants to use something for. That's fascinating. we've, been recently partnering with, this company called a Evidium, and a Evidium, is very interesting because they're, they're the type of company that is going way beyond just creating efficiency, using AI to actually be able to now offer healthcare plans ahead of time by way of very meticulously mapping medical knowledge and clinical evidence to come up with AI that is both transparent and dependable, but can now come up with options for treatments. And that's all where I think all of the people building the picks and shovels get really excited because we're like, this is why we're building all of the underlying technology necessary to enable those sort of advancements. That's awesome. time is flying. I would love to pivot to your background. What a cool company and a field you're Would you mind sharing a little bit about your story to be where you currently are? I do have a degree in computer engineering. And before Protopia, I spent 10 years at NVIDIA, And when I went to NVIDIA, my background was not yet... focused on AI, I come from a field that people refer to as computer architecture. And so many of us... would basically start learning while we were doing our job, whatever the focus was about the principles of what was going into machine learning. And learning is a very, very critical part of being able to evolve into some new space That was something that, Nvidia was very supportive of much of my focus was on applications in the machine learning space and understanding what are the bottlenecks with making these applications something that can be built out and used more broadly. And that kind of led from kind of a background in computer architecture to looking at kind of the application space of machine learning, one of which was language models. that sort of being curious about what the potential problems of this space could be was where there was a trigger for, maybe there's an opportunity for a whole business to be built. That's awesome. Well, and you know, Nvidia is a very well -respected company these days. So many people want to get in the door there. how did you get in the door for that first initial job there? Yeah, for the first initial job there, it was tied to kind of my academic background and the work that I had done during my PhD program. So I did a PhD in computer engineering at UT Austin. And the work that I had done there was very relevant to some of the problems that NVIDIA was solving back then Well, it's so interesting to have that technical acumen and then mapping it to projects of a company. I love the idea of like Nvidia, I hope many companies, I certainly do with my team, investing in your talent. Like, you know, I don't know if you've heard the phrase like, no, if we invest in them, what if they leave? oh but what if they stay? to really develop talent internally, especially as the field is changing. You want your team who might stay with you for a decade as you did to really be constantly iterating and upskilling is I think a very relatively new age thought for a company to be doing. yeah, I think it's, it's, it may be a new age thought, but it's, it's also something that I think at this point, any organization would be a miss not to consider. because it's, it's even, not necessarily about people leaving or not leaving. It's a matter of being able to get people that are part of a business to. be most fulfilled and happy while they are there. part of this evolution in AI has been creation of a lot of guidance around how to contribute to this new field. There's a lot of programs that exist that aren't even university degrees. They're like online programs or online coursework for six, nine months, maybe. to actually dive deep into what it is that you need in terms of skills to understand what's going on here from either a technical perspective or a business perspective. and then be able to contribute to that. it's just worth noting as well about this notion, since the word PhD has come up a couple of times, I really want to emphasize that that sort of formal academic training. is by no means necessary to be very effective and have very fulfilling roles, either on the technical side or not in this new ecosystem. So unless somebody got a PhD in the past five to 10 years, they didn't learn these specifics in school. They learnt it by way of being curious and using the very vast amount of resources that are available to go learn those things. Any academic training teaches people is not the actual subject matter. It's the way to think critically and the way to learn something. So that's a skillset in and of itself. And you only get that by, by practice, right? Like anything else, you have to flex the muscle of learning. That totally makes sense. If you were truly, as you mentioned, in the trenches figuring things out in a six to nine month period of intensity. There's not things you can't Well, thank you so much for sharing. If people wanna learn more about you, follow you all, check out Protopia, where do they need to go? Yeah, so our website is protopia .ai and there's plenty of information about what we do there. We have a LinkedIn social channel where we kind of keep... up to date And we'd love to hear from you if there's interest in any of what we're doing and if especially any of the listeners are in the world of business using generative AI for either developing new products or being able to drive efficiencies in their businesses. we will be happy to show you what we've built and look at potentially partnering. That's awesome. Cool. Well, thank you so much for your time and have a wonderful rest of your day. Thank you, Joan Bye bye. Oh gosh, was that fun. Did you enjoy that episode as much as I did? Well, now be sure to check out our show notes for this episode that has tons of links and resources and our guest bio, etc. Go check it out. If you're ready to dive in to personalize your AI journey, download the free Your AI Roadmap workbook at yourairoadmap .com / workbook. Well, maybe you work at a company and you're like, hey, we want to grow in data and AI and I'd love to work with you. Please schedule an intro and sync with me at Clarity AI at hireclarity .ai. We'd love to talk to you about it. My team builds custom AI solutions, digital twins, optimizations, data, fun stuff for small and medium sized businesses. Our price points start at five, six, seven, eight figures, depends on your needs, depending on your time scales, et cetera. If you liked the podcast, please support us. Can you please rate, review, subscribe, send it to your friend, DM your boss, follow wherever you get your podcasts. I certainly learned something new and I hope you did too. Next episode drops soon. Can't wait to hear another amazing expert building in AI. Talk to you soon!