These 20- and 22-year-olds raised $5M from YC, General Catalyst to study online behavior using vision AI

Amogh Chaturvedi^[1] is running on little sleep but plenty of conviction at 6 a.m. He’s groggy, apologetic for rescheduling, and still reeling from a recent scare involving a family member and an electric scooter.

Within minutes, though, the 20-year-old Stanford dropout snaps into focus, walking me through how he and his co-founders sold one startup at 19, landed in Y Combinator, and raised $5 million for their next company, Human Behavior^[2].

Launched just a few months ago, Human Behaviour is betting that vision AI can do what analytics tools like Mixpanel and PostHog have struggled with: give companies a real understanding of how people use their products, including why they convert or churn.

Instead of relying on manually tagged events or clickstream data, Human Behavior claims its AI watches real user session replays and generates insights, answering product teams’ most pressing questions without hours of instrumenting code.

The four-month-old YC startup closed its $5 million seed round in just two days (which is becoming a norm for current YC companies), with backers including General Catalyst, Paul Graham, Vercel Ventures, and Y Combinator.

“We could’ve done the financial engineering game because we got more offers with higher valuations, but we didn’t want that,” said the CEO. 

Chaturvedi met his co-founders, Skyler Ji^[3] and Chirag Kawediya^[4], both 22, at a hacker house he organized in 2023 as an excuse to live and build with friends after his freshman year at Stanford.

Their first startup, Dough, was an e-commerce accounting tool they bootstrapped. Like Chaturvedi, Ji dropped out of college (leaving Berkeley) while Kawediya went on to graduate.

Although YC was initially skeptical about Dough’s market potential, the team was admitted into the accelerator’s spring batch this year on the assumption they would eventually pivot, Chaturvedi says. They did so almost immediately, after speaking with every customer and inquiring about any other problems they faced.

The feedback was consistent: while Dough could show which products were selling or not, the customers wanted to know why. Answering that required analytics powered by behavioral data, not just accounting reports.

With this new direction, the team sold Dough for six figures to Employer.com, the same company that bought Bench^[5], and went all-in on Human Behavior.

Kawediya explains that companies using traditional analytics often need engineers to set up event trackers for every button and click, burning hours, sometimes weeks, of engineering time.

For a fast-moving startup, that’s far from ideal. “Even once you have that data, you’re still stuck with the bigger question of how users actually interact with your product so you can make it better,” he says.

Session replays aren’t new, but until recently, computer vision models weren’t accurate enough to parse them at scale. Now they are, and Human Behaviour is doing so to summarize and segment thousands of hours of footage. “Why spend hours writing code to track clicks when we can just watch the video?” Ji adds.

Today, Human Behaviour’s customers — mostly fast-moving Series A and B startups — get daily summary emails highlighting which features were used, which bugs appeared, and which users churned. Since launching four months ago, Chaturvedi says the company has been growing 20% month-over-month.

The founders call session replays an “untapped goldmine.” Right now, Human Behavior helps teams understand users and squash bugs. Over time, the same dataset could power automated QA and embedded IT support. Their ambition is to make Human Behavior the Datadog of session replay, spinning out dozens of products from the same core data.

Building with new technology from the ground up is how the founders believe they’ll take on more established players like Mixpanel and PostHog. “For some of these companies, it might be difficult to replicate what we have because their architecture can’t support the shift without starting over,” remarked Chaturvedi.