When Krish Chelikavada tried to raise money for a startup as an undergraduate at New York University, he didn’t get much traction. A few years later, after arriving at Stanford University, investors became much easier to reach.
The difference, he said, wasn’t just the business idea. It was the network.
“I think being in a place like Stanford just gives you that access,” Chelikavada said. “There’s a lot of credibility attached to that tag.”
Today, Chelikavada and co-founder Keon Kim lead Om Labs, a venture-backed startup building AI quality assurance (QA) engineers that use computer vision and browser automation to navigate websites and applications like human users. The goal is to detect problems before users encounter them.
The startup’s journey offers a glimpse into how Stanford’s entrepreneurial ecosystem continues producing AI startups at remarkable speed.
The founders closed a roughly $1.9 million pre-seed financing round in 2023 from investors including Alliance, Balaji Srinivasan, Soma Capital, Cory Levy’s Z Fellows and Hashed. At the time, however, Om Labs was not an AI company. It was a crypto startup that managed customers’ cryptographic keys.
The founders had spent two years building distributed key management software for Web3 applications after spotting what they believed was a major technology inflection point. Then generative AI arrived.
Instead of stubbornly sticking with their original business, they did something many founders struggle to do: They started over.
“The whole reason we raise VC money is to go after the biggest opportunity,” Chelikavada said. “That sometimes requires you to ‘kill your babies.'”
The road to VC funding
The decision reflected lessons Chelikavada says he absorbed at Stanford.
One influential course was taught by venture capitalist Ann Miura-Ko of Floodgate, who encouraged students to recognize technological “inflection points” that create opportunities for entirely new companies. Web3 appeared to be one such moment. Soon afterward, generative AI looked even bigger.
Chelikavada made a conscious decision to pursue entrepreneurship, even declining a job offer from Microsoft. His rationale: “If this works, great. If it doesn’t, I’m still young enough to go back to the industry.”
Chelikavada studied computer science before earning a master’s degree in Stanford’s Management Science and Engineering program, where he focused on product design and behavior science. Kim brought complementary experience. He spent several years at Uber building machine learning systems and wrote two books on machine learning that became popular in South Korea.
Even with Stanford’s reputation, funding did not happen automatically.
Chelikavada said one of the first breakthroughs came after applying to Cory Levy’s Z Fellows program. Stanford professors, alumni and visiting venture capitalists opened additional introductions, eventually creating the network that led to Om Labs’ financing.
“You have to kind of play the network a little bit,” he said. “You need access.”
He argues entrepreneurs generally earn that access in one of two ways: by building credibility through work or education, or by placing themselves inside startup ecosystems such as Silicon Valley where investors, founders and engineers regularly intersect.
Choosing the right product
Finding the right business proved harder than raising the money.
After pivoting into AI, the founders spent several months testing ideas. Their first concept — an AI customer-support agent — quickly attracted interest. They abandoned it anyway after concluding the market had become too crowded.
Instead, they worked backward from a broader vision. “Let’s start with a vision for what we think the world looks like, and then work backwards,” Chelikavada said. “The theme that we were personally very interested in was AI employees, but specifically RPA, which is robotic process automation.”
RPA consists of software programs that perform repetitive and rigid business tasks. For example, one task could be opening a company’s email, downloading invoice PDFs, reading fixed fields (for example, invoice number, vendor amount), entering the data into an ERP system such as SAP, routing invoices over $10,000 to a manager for approval, then marking the email as processed.
But if a form is redesigned, a button moves or there are other changes, the automation breaks because it isn’t flexible enough to adapt. In contrast, AI agents can reason about what they see. If a format changes, it could still recognize the data and continue the workflow.
“RPA is traditionally very deterministic, and AI can actually flip that for you; you can add a level of reasoning. QA testing is one classic RPA use case. So we got to QA testing that way,” he said. With the belief that “AI-powered RPA is going to be huge in the future,” their task became, “how can we reduce that to a vision that we would enjoy building today?”
The result was their AI QA engineer product, Jina. It watches screens, clicks buttons and verifies whether applications actually function correctly.
“Jina basically uses browser agents to simulate how users use your app … to test new features and catch any issues before users have to face them,” Chelikavada said.
More than 250 companies have signed up for early access, according to the startup, with Om Labs gradually onboarding customers while refining the product. The startup is a two-person team, but Chelikavada said they do the work of five to 10 people by relying heavily on AI.
Staying ahead of competition
The two founders thought hard about what would make their product stand out. They landed on the idea of context.
“Today, coding agents basically search through files like humans do,” he said. “What we’re trying to do is build this graph of dependencies that even before the agent looks at it, we have a map of how every system is connected together – what function calls what function, what services are tied to another.”
As such, “whenever you make a change to the system, we know all of the paths that are impacted by that change. Then we run agents on those paths to try and identify issues.”
But couldn’t a larger AI company replicate what they do to put them out of business overnight? “There’s a lot of fear among startups that OpenAI and Anthropic … they’ll eat everything,” Chelikavada said. “I’ve kind of realized over time that’s totally not the case. Startups can still win by focusing on one use case that leads to a bigger vision. So the laws of business have not changed. Focus can still win.”
That’s because the OpenAIs and Anthropics of the world are not interested in pursuing every market niche. “I don’t think they can fight all frontiers,” he said. “They’re just not incentivized to do that.”
In the meantime, they will continue to build Om Labs from the right starting line: Stanford.
“I think it’s honestly being in the right place at the right time, being around the right people and building enough credibility and making those connections,” he said. “Because I tried this as an undergrad and it didn’t work for me.”
Be First to Comment