The Wunderkind Who Won Over NASA, Stanford and Peter Thiel

Art by Mike Sullivan. Photo George Sivulka.

George Sivulka on cold-calling scientists at 15, shaking down his college for more units, and raising $30 million for his AI-powered search engine.

When George Sivulka was 12, he was building lasers that could light fires. At 16 he had an internship at NASA. By 22, he was hanging out with Peter Thiel, talking for hours about artificial intelligence and philosophy. Thiel, thoroughly impressed, wrote one of the first checks for Hebbia, Sivulka’s AI-powered search engine. Instead of relying on keywords, Hebbia, founded in 2020, can process more-complex requests. Sivulka gives the example of someone in private equity sifting through hundreds of pages of company filings to find a business’s risk factors; Hebbia could analyze the document and return risk factors, even if the answers didn’t contain the user’s keywords. Investors are banking on the prodigy, with Hebbia raising $30 million in a July Series A led by Index Ventures. Now 24 years old, Sivulka walks us through his journey from selling refurbished computers on eBay to raising millions for his startup.

Sivulka grew up as a first-generation American in New York with a Slovakian father and an Italian mother. He had an innate passion for math, and by the time he was 12 years old, he was digging into archived programming forums from the early 2000s and scouring the streets for electronics.
I would pick up computers and projectors in the middle of Washington Heights and take them home. I’d extract the diodes and build lasers—I could light things on fire with them. I would reassemble computers from old parts and sell them on eBay. I’d find one for free and I’d sell it for a thousand bucks.

On a snow day when he was 15, Sivulka walked out of his home and trudged to the NASA Goddard Institute for Space Studies at Columbia University, in search of a job.
I printed out my resume on really nice paper and put on the only suit I had —it was a pretty ratty suit—and I walked down to NASA’s office. I knocked on the door and the security guard kicked me out. Then I went outside and I googled everyone’s names and phone numbers. I called every single person in the building. Finally, the director of the lab answered. I said, “Hey, I printed my résumé and I’m standing outside and they won’t let me in. Can you please come down?” I had a two-hour conversation with him in front of the security guard that kicked me out. I convinced him to give me an interview.

He spent that summer as a NASA intern alongside some of the country’s most brilliant scientists, helping Dr. Reginald Eze, a mathematics professor at LaGuardia Community College, with his research into programming satellites to detect land mines.
That summer was like a crash course in calculus and physics. I was learning an insane amount. And by the end of it, I contributed to research [that was submitted to the 2015 Comsol Conference]. We traveled to Boston—I was all of a sudden wearing a nicer suit—and we actually ended up winning [a major research award] at this international conference. I didn’t even know the magnitude of it at the time, but it was a pretty big deal. I remember my professor telling me, “You want to go to Stanford, right? I think you’ll have a good shot now.”

Dr. Eze was right. In 2016, at the age of 18, Sivulka left New York and enrolled as a mathematics student at Stanford University. Upon arrival, his ambition chafed against some of the college’s academic rules.
I petitioned the administration 10 times until they let me take more units than was possible—more units [than] almost anyone has ever taken as an undergraduate. I matriculated into a master’s as quickly as I could to allow me to take even more classes.

Sivulka finished his undergraduate degree in two and a half years. He went on to do a master’s program in applied physics and then a doctorate in computational neuroscience. His graduate studies took him both deep into the brain and to the bottom of the ocean.
I became really, really interested in machine learning as applied to computational neuroscience—modeling the brain and coming up with machines that were intelligent and could help humans. I was training machines to behave like the earliest coral brains. I was free diving every day collecting corals in Micronesia. I was actually the first person to record action potentials in five different coral species. It was one of those experiences where you feel like you’re probing into the unknown and you’re the first person to see something. It’s an indelible experience. It stays with you.

As Sivulka was diving into coral reefs, his friends were graduating and starting careers in finance and consulting. He heard stories of their 80-hour weeks and mountains of busywork.
You have the most intelligent people coming out of Stanford and all of these amazing schools, and the majority of them go to jobs where they’re just rendered automatons. It’s a new kind of brain drain where you have smart people do stupid, mundane work. When I looked at the people that were graduating, they hadn’t smiled in days.

At the same time, Sivulka was learning more about how far AI reading and writing models had come. Sivulka thought he might be able to use this new technology to help his corporate cog friends.
As soon as I realized that machines were picking up these abilities to read and write, I started doing research into these new kinds of transcendent models. I built a demo of [Hebbia] that you could ask as if an expert had read the document and knew where all the answers were instantaneously. I remember my friend in investment banking would spend all day reading this one kind of [Securities and Exchange Commission] document. It’s this 400-page document and you’re just really looking for a few different questions. [Hebbia] saved her like three or four hours. I remember her face [when she used it] being a bit alarmed. When you see a computer do the work for you and make you that much better at your job, it’s an “aha” moment.

By this time, Sivulka had impressed quite a few people in the tech world. A few friends passed his name and his idea for Hebbia to Thiel.
They told me explicitly that Peter Thiel wasn’t investing, but that he’d be willing to have an advisory chat with me. I drove [from the San Francisco Bay Area] all the way down to his house in Los Angeles. He had a lot of compassion and empathy, and it was a wildly interesting discussion. I felt like I had made a friend. We had an incredibly long conversation and he offered us a [pre-seed check] at the end of it.

Hebbia was one of the very few startups to ever receive a pre-seed check from Thiel, joining the likes of OpenAI and Anduril. That sent a strong message to Silicon Valley: Hebbia was a company to watch, and Sivulka was a founder to bet on. With a $30 million war chest, Sivulka plans to double Hebbia’s 15-person team in the next year and keep building an AI product that will transform industries like finance and academia.
There’s too many [AI] applications right now that take time away from people. Netflix sends you down a movie rabbit hole; Twitter increases political polarization; all of these Amazon recommendation systems feed you more items to buy. None of them actually give time back to the user. Hebbia recommends the right answer for you and shows you exactly what you need for work, whether it’s in a really large filing or some academic research paper. I don’t want humans to be replaced. I want them to be empowered.

Margaux MacColl is a reporter for The Information, covering tech culture. Previously, she was a startups/venture capital reporter for Business Insider.