Suman Kanuganti of Personal.ai: AI-Enabled Messaging That’s Personally Yours

Startups get the party. Startups get the party started. Startups get the party. Startups get the party started. You want tech and tacos, Tacos and tech you know we got tacos and tech is on smash border green guacamole keeper captains counting beans Contador quentan frijoles found us trying to raise some queso la vida San Diego school of rappy rapping fire tortillas round jalapenos in these modern times a crappy my coding yoga posse fold the body origami we be surfing coronado eating tacos el gordo the tacos standing puesto lucha libre and sombrero got a shout out to cost angels by the ocean watch the sunset La Jolla shores Entrepreneurs trying to get funded When San Diego Startup Week is popping in the summer Oceanside, the Tijuana Estee be full of hustlers from couch surfing to penthouses right on the beach Tech leaders believe stars are within reach we interview and founders who be shooting for the moon SD Tecos with your host, Neil Bloom. Startups get the party. Startups get the party started. Startups get the party. Startups get the party started.

Welcome to the Tacos and Tech Podcast. I'm your host, Neil Bloom, and I'm thrilled to have you here. On the Tacos and Tech Podcast, we interview builders of technologies, of companies, of communities, and we do it weekly, so thanks for tuning in. Every Tuesday, a new interview and on Fridays we talk about what's new in the San Diego tech ecosystem from a news perspective. So make sure you join us there as well too. We also have a tech newsletter that comes out every other week@freshbroodtech.com you can subscribe to listen to it without further ado. Enjoy this episode and welcome back to the Tacos and Tech Podcast. I'm your host, Neil Bloom and I'm thrilled to have on serial entrepreneur and co founder and CEO of personal AI Suman Kanaganti. Hey, Suman, thanks for being on the pod.

Hey, Neil, always. Pleasure.

It's a pleasure. Is all ours to get to know a bit more of the story. We've actually had your co founders on almost two years ago, maybe a year and a half ago.

I think it's like. Yeah, around 18 months ago. Yeah, yeah.

At a different iteration of the name and I'm guessing a different iteration of the company. So I'm really excited to get to know the story recently, but also your story because I know you have been focused on AI in a variety of industries for quite a while. Do you want to give us a quick overview of what personal AI is?

Personal AI is essentially a natural language model of an individual person that you can have a conversation with or other people can have a conversation with.

Okay, well, this is a very poignant conversation we're going to have based on the craziness that ChatGPT has released into the world and interest for AI at the consumer level. So we're going to, we're going to have a lot of fun talking about AI and how you're building it. Also talk a little about, you know, what you think the relevant talent pool is locally in San Diego for this, since you are building this in San Diego. But where did it all start for you? Have you always worked on kind of technical products like this? Where does it begin, your story?

I mean I always say my passion always has been solving like picking like hard problems and figuring out what technology can solve it. Over the years after working at major different corporations, I guess I picked up this desire to create more and more product experiences that are like really close to humans in terms of what they touch and feel and experience live in. My previous company, it's called Aira A I R A A also a San Diego company. And that premise was around how can you use technology to fill the gap of missing visual information for blind and low vision people and had a mix of using wearable technologies, AI as well as human in the loop to solve the problem. And with personal AI, it's kind of a similar philosophy in that how can you use technology to augment your own mind and memories that otherwise are lost and forgotten? What does it mean to have the full access to you for not only yourself but also people around you. So that's the genesis and the premise of who I am and what I am passionate about building.

Okay, incredible. Obviously really you're dedicated to creating a pretty big impact on the world using technology. Do you come from a family of other technologists or entrepreneurs or how are you surrounded growing up to kind of know this is where you wanted to do it?

I don't know, man. I think it's in the blood because I grew up in India and in India you either have to become an engineer or a doctor to make a living. That's what your parents taught you. But no, generally, even I think personally as well, I enjoyed the engineering and I enjoyed science for a long period of time. I came to United States with this desire to continue education. I did my robotics because I was super interested in robots like back then. And then I picked up software Hardware systems engineering. And there was a point in my life I wasn't into it as well. You know, speaking of San Diego, I was in San Diego for 15 years now. Yeah. And I think there was a point in life. You know what, I know a lot about technology all the way from like the hardware to the software to the products and experiences. Maybe I can do something about it. But I have like zero clue around, you know, business and the marketing generally. So to basically break that barrier, I, you know, went and joined various school of management again here at ucsd. I think that gave me a good exposure to people, to the concepts, to just generally the business. I know people like, you know, always and then why do business just go start a company and do that shit. But I needed that lift. I needed to break in the barrier of what does it mean to have the frameworks and basis for building a company and starting a company?

Well, it was really smart to surround yourself with other technical, but business minded people or technical people who wanted to learn the business side because it's usually, you know, a gap within a company or even the gap that stops people from jumping off the ledge and starting.

Right, right. I mean, where do you use that? Like what does it mean creating a LLC to a C card? Right. Like what are the choices that I'm making? Yeah, of course we can, you know, learn through I guess creating the company in itself. But I think that time it was clear that I needed a little bit of framework to joining a MBS school and then immediately started a company within three months.

Really? That quickly? Wow. Yeah.

Amazing.

Do you remember that feeling when you took the leap and ripped off the band aid and we're all in for the first company?

Yeah, no, I do. In the first company I was, it was a wallet company. So it's essentially like a Apple pay these days. But back then we had this mechanism to convert any of your credit cards and all of your gift cards into one mastercard if you will, that you can use at any places to maximize your rewards and also reduce the pain of carrying all these cards. But we soon enough realized we were violating a bunch of different laws. And coming from a fintech background, because I was at India for a long time, I was passionate about, you know, generally the payments in the consumer industry. So that was only like one year. But the initial feeling I do remember of, you know, the desire to change the world and I think I still have it.

Absolutely. It's addicting as an entrepreneur to want to solve people's problems. Right. What was the With Aira. Tell us, Ira, tell us a little bit about the premise of that. Like where was the original concept versus where what it ended up serving? I'm curious a little bit how that the original founding idea versus where it changed to what it is today.

Ira, original idea started about with a friend of mine who happens to be blind. And the idea was around Google glasses back in 2013 or 2014, experimental basis. And it was almost. That was the one where it was not evolved like over a period of time. That was almost like a overnight thing where. Huh, Hang on. This camera could be useful for Matt Brock, who is my friend who cannot see and serve as the purpose of seeing. Right? So there was a genesis of it because it has a camera right in front of your eye and you know how awesome it would be if you can, you know, use that camera and process the information as a visual clues, you know, back to the blind person. Because technically they have visual challenges. Right. So that was the genesis and premise and we did like a bunch of different experimentations and went to meet several different blind people in San Diego and organizations and make good friends. And through that process I met a fine gentleman. His name is Larry Bock. And I think he changed my life in a pretty dramatic way in regards to who I am, what I am. He's also legally blind, so he invested, he co founded the company with me, Ira, and taught me a lot and made me who I am today. He's also important because he passed away within 18 years. Sorry, 18 months. Like working with him because of pancreatic cancer. That is the genesis story of personal AI idea. Personally, I did not happen overnight. It started off like back in 2018, where there is this constant desire to be able to access Larry's mind because it's not just emotional connection, it's also about intellectual connection that I had with him in regards to the terms, the negotiation, the strategies. We spent a lot of time together before he passed away. So okay, what does it mean to be able to create somebody's mind? To be able to interact? Yeah, sure. You know, people pass away, so there is that need. But the access problem to other people's minds exists because of status, because of time limitations, right? Because of a number of different limitations. So yeah, so that's the genesis of it, which is how can, you know, if Google solve the world's information, how can personal AI solve the access between people and people's minds? Right. So anyway, so the side of story and kind of leads into personal AI because they're like intertwined and connected for me.

Incredible. What an empowering, personal, you know, motivator and driver for you to build both of these companies. I'm sorry to hear about Larry, but it sounds like obviously he's got such a lasting impact on you and the world. And what a legacy of building a company that says, you know, hey, here's what we wish we had with him that we hopefully can change for humanity.

Larry Bob is also a co founder of Illumina. Right. You know that, right?

Great. Yes, obviously already. He had already had a big impact on the world too.

Yeah, for sure.

How did it from 2018 to even now, what has been the human AI or the personal AI story of how did you chunk off? Like what to tackle technically versus the business piece?

Yeah, it was a process for sure for personal AI because it was not a. Oh, I know the solution now. Right? First of all, even understanding the problem took a while. The problem of like memory and how people retain recall, you know, we think. So there is no equivalent of thinking in the space of like technology, right? Like we always has to take a explicit action, you know, using our keyboards or using our voice to be able to interact with technology. So the core premise was blurry. Like how can you. Because one of the desires is you need to create this experience that is natural to people. Like even with Aira, one of the reasons why the experience is so awesome is we did not try to solve for giving instructions, going from point A to point B, but naturally providing visual information and description. So blind people can negotiate with the physical environment by themselves. Like that's the most natural thing to do, which is like, how can you negotiate with things by yourself? So now in case of personal AI, how can we create this technology to negotiate with your AI or have a conversation with your AI much more naturally, much more ambiently. So earlier on, when we started thinking about the concept of creating a digital version of oneself, there were things on like experiences such as, you know, creating avatars, creating bots, creating robots. But that wasn't the essence. The essence was being able to access information that is otherwise forgotten for yourself or you know, fetch that information from other people. So it turned out to be more focused on like memories. And the problem has been that 80% creation and consumption that we people do on a day to day basis is lost and is forgotten. So we need to fill that gap. So that was the genesis of putting all the pieces together. And I know it sounds like so obvious like right now, but somehow it took a while to get to that point. So Then went into a lot of research and as I was exiting aira, which was also pretty public story, I think everybody knows about it, I started working on trying to collect what makes a personal AI and did like a bunch of different research. And earlier on we went with the name Luther AI, I think, which is kind of unit a little bit. And the story back then was kind of similar, but the inspiration was like being able to express yourself with this core idea of capturing speech as a information source. Right. It's like, you know, what you say, what you hear is very powerful. And one of the key information source we started off was conversations, was speech and talking. And there is like this sense of like express yourself without being able to forget and remember things and remember people for a long period of time. So I wrote a blog about why it's called Luther AI. But later on, it has evolved so much that speech is not the only source for making up people's mind and conversations. It has other implications as well, like other sources, the things that you write, the things that you see. Right. So we evolved the vision to a much broader, which is personal AI today with this core idea that it doesn't matter what kind of modality or information exists. It is a memory and that is comprised of context, people, location, images, everything else. So, yeah, as of today, the memory stack that we talk about represents a digital memory world of an individual person. And it has this automated ways to take the unstructured data to structured data that then goes into your own national language model, which is a technical LGBT architecture implementation. And rest of things is fast. Because that's three years of work.

Yeah, absolutely. And I think even in the timeframe of where AI has advanced, it's pretty quick for what you guys have advanced it too. So we'll talk a little about the technology advancement, but I want to touch back on something you mentioned about, you know, you said a very public exit from aira. What are some things that you learned from company one to company two now, like anything you share with brand new founders that's like, hey, do things differently or here's what I did, you know, anything like that kind of come to mind on company number two?

Yeah, I think there are many. But let me try to summarize some of the things that personally I practice myself. There are obviously things that I double down on and there are things that I kind of step back and upload things differently. The things that I double down on are people. Like, I think there is nothing more important than focusing on solving for people and staying Grounded. We can fall in love with the technology, we can fall in love with the fame, we can fall in love with the money. But the most effective and the compounded thing that I think for me personally is being able to have this ultra focus on the problem and the people and rest of it doesn't matter, right? There are always ups and downs. It's. I think nothing is clear. Like, you know, in startup world there's always going to be tough times and there's going to be hard times. People are going to be good to you, people are going to be bad to you. But at the end of the day you had to have your own choices on how to balance and stay grounded through those times and focus on what the problem is and focus on whom you are serving. Like customer focused, customer social. So that I think I still carry that through this company. And personal AI is so broad, right. Like if think about it generally to everybody, right. So that's probably the second one which is Aira was so impactful, but it is impactful to a small group of people which is blind and low vision population. And one of the challenges is like understanding like the size of the market itself and it was like a niche market. So it does make a good business, but it's a lifestyle business. So the alterations of second, like pivot, if you will, in the second company is how can I serve like a, you know, broader group of people in a much more effective way. So rather than choosing a niche, it's not a bad thing. And in fact there are problems with both side of the bodies. Meaning if Aira was a niche, it's very, very clear on the market, very clear on like what the experience is, the business and everything else. But then what are the expansion plans? Right? So there is that problem in personal AI. The problem is reverse which like, oh, hang on, oh. So who is your first niche? Like where are you going out of what? You know, which problem are you solving? I mean memory, everybody has memory problems, right? So. So I think it's more about like the go to market strategy and the problem and differences. So. So I think it's just the constant desire of staying true to the vision of the company that you established. And with Aira the focus was very much on blinded low vision. And that's the reason why it was never a expansion into personal AI. Like Aira has to be Aira. That is the mission for Ayra. Personal AI mission is different and it is serving everybody. What else? I think on the fundraising side, one of the things that I Changed is especially like as I go from like you know, pre C to CDC to series B. I think the strategy now is rather than taking like any people's money, any business money, I think if you are in the places where you put yourself on what the problem you're solving and how you are solving it and you know, start creating, showing, showcasing those experiences. Choosing the right investors who believe in the mission in you for the long run, I think it's much more important than trying to really, you know, focusing on immediate milestone that you may need to achieve. And luckily with Ivra, I got those investors early on because of Larry Bock and Lux Capital. And the strategy now is, you know, choosing the right set of investors for the needs that you have for the company to build it in the long run. And it's not like a short term return.

Really insightful and I think we're told this as founders and it's hard for sometimes some of this to hit home until you hit it right in the face and be like, oh, learning moment. So pretty insightful. Pretty great advice. I think it's actually really insightful the way you talk about kind of total market in a sense, right? Like the niche versus broad to everyone, but then having a niche back that I found that really poignant around that. And that comes up in a variety of ways. Like sure, maybe it comes up in the pitch to the VC that's asking those questions. But then you realize like, oh, actually who are we building this for? Who's our customer? Can we really build the Persona and get to know their every need for what they want and then solve that problem? So that's awesome to hear those learnings and I think our audience will really, really connect with that. Where's personal AI now? Like take us through a little bit of the technology development and the product. Like what can people interact with today?

Personal AI today has developed something called a personal language model. We developed it like almost two years ago, but there were a lot of fine tuning and updates to the language model that was happening. The language model itself essentially supports three things which we call Personal Q and a model. Essentially being able to ask any questions, open ended questions to yourself and other people asking questions to yourself. The second level is personal generator model. It's essentially a GPT architecture implementation. If you know GPT3, which is a large language model that is trained on Internet of data. We've been coding GPT architecture that is much more focused or tuned or designed for personal language model with around like 120 million parameters with Individual data. Right. This is a big subtle difference. It's kind of almost opposite spectrums. So we have opened up personal AI 1.0, we call it to beta, like last year. And because, because of how the market was last year, we started off with taking subscriptions to begin with because it does cost us, you know, GPU money and container money to be able to host. However, things have evolved to a point where we optimize the models to, you know, being able to extract information and recall information, even with small amounts of data. So it's actually very beautiful. If it goes outside the bounds of your corpus, which is your memory stack that I was referring to earlier, then it falls back into leveraging the general models, such as GPT3, ChatGPT, if you will. But the architecture is designed to use your stack, your model to extract information and understand the intent and respond back. And if it does go outside the boundaries, then it kind of leverages GPT3 and ChatGPT. So that's essentially where we are. We are now, as you know, when this podcast is released, personal AI is available for everybody for free. So you can technically just simply come in. If you sign up to personal AI, you will get your AI. It's more or less think about it as a chat application or connecting with other people like a WhatsApp or like a signal or like a Slack. Except that in personal AI, as you have these conversations with other people or with yourself, which is your AI, it will constantly learn and train in any context that it understands. It provides suggestions to be able to recall for you, construct the response for you in the manner that intent is understood, and you can simply swipe right to send or swipe left to edit and personalize it further. And then again, it goes back to the learning process. So it's a system you can obviously bootstrap much faster by integrating, let's say all your Twitter feed, integrating all your documents or knowledge that you may have behind the scenes. So you can do that. But at the core, the goal is to evolve this model to be like you stylistically, intellectually, emotionally. And it's an asset that always is growing. It's not a model that you build and then put it on the website. It's continuously training as life evolves.

That's great. Well, congrats on the release and the launch. I feel like I think I've had a signed up account since the Luther days and so I've appreciated the insights from you that you've shared in your newsletters over time. I've got to imagine with just like the hype that's happened locally with ChatGPT, that that's a good thing for you in the industry, right? Have you felt, I mean, you've been focused on AI for a while, but have you felt just the floor kind of raise and the whole community and everyone just talking about AI a lot more in a different way now?

Oh yeah. I mean, thanks to ChatGPT because it's almost very interesting because back in November December, we had like a few thousands of beta users in personal AI. And personal AI has a personal AI chat. Technically you chat with your AI, right? It's a chat interface. That's basically what we built. And luckily enough ChatGPT also released chat interface. So in a way it validated that we were working on the right things and it was not something totally off, like totally off on interactions and the experience. And I do remember even like six months ago, one year ago, people would go, this is crazy. We will see when it happens. I don't know if the AI capabilities would be there. I don't know if consumers would actually adapt to it. And luckily because of ChatGPT, it gave an awareness on terms of what the AI can actually do. And I believe, I mean, not I believe I also know that personal AI is perfectly positioned because it's OpenAI is a large language model. It's akin to Google search, you know, for you to essentially get the information from the Internet much faster and get that insights from the Internet much faster. Yes, you can push in some data and summarize things, but at the end of the day it is a large language model. And our goal is personal AI, which is like a personal language model. And the beauty is the chat interface is similar. How you choose the intents and write the intents is very similar because it is same architecture at the end of the day, which is GPT architecture. So if you are learning ChatGPT, then the learning of a personal AI is much faster now. So why would I not be happy? So I am super happy. And people want personal AIs, right? And we are extreme in regards to the personalization. You know, I do see like a lot of people like getting creative, using all the organizational knowledge and you know, pushing it to ChatGPT. But the thing to remember is, you know, every contributions in every data that is going to ChatGPT does go into LLMs, right? So it will be indexed in the future. Like when you have GPT4, it is likely that your information will be indexed as well. So I think there are Use cases and there is market for OpenAI, as we all know, and for charging. But our focus is purely human. The individual AIs, individual models to help them, you know, recall who they are, what they are, what they want to remember and help them connect with other people in a way that is so personal that anything north of 120 million parameters, or even 200 million parameters, it won't be personal enough. Right? That's the technical thing. Yeah. And you know, generally AI, I think there is good visibility into the large language models and we all know how much compute it requires, like you know, 10,000 north clusters of GPUs. So there is a huge contention on the GPU, you know, supply and big tech is fighting for that supply. So anyway, hardware is behind a little bit in the AI space, but for us it's actually an advantage because this 120 million parameters is so small that our eventual goal is to run on edge on your computer, on your mobile phones. So in a way we scale horizontally. We are not dependent on the fight that is happening. And there's also attribution problem which is in the case of personal AI, you own the data. Every AI response is attributed back to the response or back to the memory stack data. Right. So you know, who the hell is the owner of that particular data. And there is not a confusion on ChatGPT has created the response and who needs to get credit for it? Well, it is coming from the Internet, right. It has no attribution. So I think from a individual level, I think the missions are different, but at the same time it is AI. The architecture is very similar, the capabilities are similar, the interactions are similar, which is good. But at the same time it creates like confusion on what the hell is personal AI vs what the hell is OpenAI. So as long as we focus close to the people with the angle of privacy, with the ownership, with angle of their model, their data, this is used to connect with yourself and to connect with other people. That's where, you know, I would stay, I would stay with humans.

Got it. Really, really interesting. I think just like a lot of other, I don't know, buzzwords in tech, what really will, what you are driving for is the eventuality that who cares if we're calling it an AI. This is going to be, this is going to be built into just the way of life, right? And you're going to have a multitude of them. You're going to have your personal, maybe you're going to have your work, maybe you're going to have you know aspects of that people and I mentioned the buzzwords only that people talk about crypto in the same way. Crypto will be crypto when we're not talking about crypto and it's just happening behind the scenes by using blockchain technology for the things it needs to do. Right. And I think that's what you're driving at here is that you know, when you're able to data mine yourself that success to a certain regard and you guys are well on your way there. I'm really, really excited for that. What should we be looking for, you know, a year from now? You think, how are we going to.

Be using our personal AI from a year from now? I think it should be more integrated into your daily communications. In other words, like every communication that you are having will have a personal AI attached to it if you would wish to. And that means that you would never miss a moment and you never forget a moment and you would start creating an asset for yourself. It's almost like investment is how I think about it. It's like investing in yourself. And the beauty in here is most of the assets like depreciate over time. This asset appreciates over time because it's not only growing, but it's also constantly evolving and learning. Meaning models only will get better, right? If the models are so good right now, it will only get better. So eventually the trust that established between you and your AI will increase. That means that trust will carry to people around you. I have some interactions, you know, I'm testing my. I think we are recording this in Feb right now we are in beta. And when people talk to me, when my AI says something and it follows a personal score such as like 63%, you know, they trust it already. I still go ahead and you know, give my blue thumbs up, print in that on those messages. It's not blocking, it's a continuous connection, you know, that people are establishing with me. And I do the same thing with Sharon. Like, you know, in middle of the night if I had to like write a report or board report regarding getting. How she talks about, you know, generating grounded transformer which is gt, you know, akin to GPT. Basically fetch those information and I don't have to disturb her and call her, hey, can you give me this information? So I think the world will become more continuous and it will be more natural. And the key for us is an integrated experience. So it's not a separate thing. It's not a bot that you go and talk to. You talk to humans and AI Is your extension, you know, part of your extension, and it's. It's integrated into your life. Got it.

Great. Well, we want people to go and try it right now and give feedback. Right. So where can people go to engage now? As of this release, that's been out.

Yeah, Yeah. I think when people are listening to this. So you can create a free account, simply personal AI. You sign up and your AI will initiate a small model that you can use to interact with other people. You know, invite people, create lounges, create groups. It's no different than a WhatsApp. The big difference, though, is you will have your own model that continuously will evolve.

Okay, well, I'm very excited to dive in personally into personal AI and dive in and be a user as well, too. You've been in San Diego quite a while. You're dedicated to building more companies here. You obviously then have interacted with our taco scene. Do you have a favorite taco scene? Do you give yourself notes on tacos in your personal AI as you. As you taste?

Yeah. You should probably go ask my AI, what is my favorite restaurant? The funny thing is, even if you ask my favorite restaurant, it will give the same answer. Which is Taco Place, and that one is Las Hadas in downtown. The owner, Alan, is a good friend of mine. We lived in downtown for a long time. My wife, it's on there 4th and Market. So I love the place. The tacos are good, the people are good, People are good, I think. So I go there for experience. I go there for, you know, catching up with my friends. But I also enjoy, like, really good tacos there. And their favorite dish is called, like, Enchilada del Mar. It's really good too. But my other favorite taco place for online orders, even, you know, during lunch is Spitfire tacos. Spitfire tacos is good too. It's something new. They have a little bit of, like, different touch to tacos, which I kind of always enjoy, like fusion in style, which is kind of my style of cooking as well. Anyway, so, yeah, so those are my two favorite taco places. Awesome.

I love it. Excited to add Spitfire as a new entry to the taco database. Great spots and good dishes. So Suman, really appreciate the time and the energy you're putting into to building this personal AI. So thank you and thanks again for being on the podcast.

Thank you, Neil. Talk to you soon. Cheers.

Thanks for tuning in to the tacos and tech podcast.

You can find all of our episodes.

On freshbroodtech.com as well as on Twitter at Tacos and Tech Pod. Also, follow Fresh Brewtech on all of our social platforms and make sure to share and give a review on Apple Podcast, Spotify and Google Play.

Thanks again.

Party started, you want tech and tacos, tacos in tech you know we got it. Hey startups get the party Startups get the party started Startups get the party Startups get the party started Startups get the party Startups get the party started you want tech and tacos, tacos in tech, you know we got it.

Hey sa.