Stay up to date on upcoming events, deadlines, news, and more by signing up for our newsletters!
For Denis Yarats (Courant), what began as a Ph.D. project at NYU became the foundation of one of today’s fastest-growing AI startups.
“Sometimes it feels incredibly impossible, but you just do it one day at a time and see what happens,” he said.
Driven by the belief that knowledge should be accurate and accessible, Yarats built Perplexity, an “answer engine” that aims to bridge the gap between academic rigor and curiosity in an age of information overload.
From Papers to a Company
During the 2020 pandemic lockdowns, Yarats co-authored a research paper on reinforcement learning from human feedback. Just two days later, Aravind Srinivas, then a Ph.D. student at Berkeley, published a paper with an almost identical idea.
“It’s funny that’s how we started talking and collaborating,” Yarats recalled. “For the average person, the inflection point was the ChatGPT moment. But it was visible years earlier for researchers.”
Their friendship blossomed as they exchanged ideas throughout the pandemic and stayed in touch even as Srinivas joined OpenAI.
By the summer of 2022, they’d decided to pursue their own opportunity and formally launched Perplexity just a week after ChatGPT.
Answer Engine That Comes with Citation
For Yarats, Perplexity stands out in a crowded AI startup market because of its commitment to research. Just like how your professors constantly remind you to jot down citations as you go, he believes facts should be traceable. The startup says every answer generated comes with verifiable references you can click and explore, limiting the risk of “hallucination.”
The company’s name reflects this commitment. “Perplexity is what we worked on every day as Ph.D. students,” Yarats explained. “It’s the synonym of entropy, the uncertainty you’re always trying to reduce.”
Academia and Startup Culture
Some might say Perplexity has adopted the structure of a traditional tech startup, but behind closed doors, it still has the backbone of a lab. By limiting each project to no more than five people, the team can run experiments and tests quickly instead of getting bogged down in coordination and delegation.
“It feels like I’m still doing research,” he said. “Now I just have the resources to pursue bigger, harder problems.”
Looking Ahead
Earlier this month, Perplexity announced the global launch of a new AI-powered web browser named Comet. Through its conversational interface, users get an experience close to having a personal assistant who can automate tasks from organizing tabs and drafting emails to shopping online.
“It feels great when you have an idea, you test it, and then the rest of the world agrees with you," Yarats said, marking an important step in the company’s goal of unlocking a new distribution channel and reaching an even wider customer base.
