The recognition arrived before the company turned two. In the spring of 2026, the firm Aspire named Saarth Shah one of its 30 global founders to watch, a nod less to the size of his startup than to the problem he had chosen to tackle.
Shah, who was born in India and studied data science at the University of California, Berkeley after transferring from the University of California, San Diego, has spent most of his short career circling a single question. How do you make a computer’s answer about the real world trustworthy enough to act on?
He took it seriously as an undergraduate. At the San Diego Supercomputer Center, he built geospatial models that read abnormal 911 call patterns to alert authorities to mass-casualty events such as school shootings. The stakes taught him early what accuracy costs. “One wrong alert, and police and school authorities mobilise for something that turns out to be nothing,” he said. At Stanford’s Snyder Lab, the genetics group run by Michael Snyder, he helped build Wearipedia, an open-source toolkit that makes data from wearable health sensors usable for students and smaller labs. The two efforts shared little except the discipline underneath them, deciding how far a dataset can be trusted before someone relies on it.
He was building companies the whole time. As an undergraduate, he co-founded Socale, a campus networking app backed by Berkeley SkyDeck and Blackstone LaunchPad that used a graph-based recommendation engine to match students based on academic interests. It passed a couple thousand downloads before he moved on. In 2023, he co-founded Dart, where he built a conversational voice AI receptionist for medical clinics on top of GPT-3.5, Twilio, and Deepgram, a system that answered calls and booked appointments by phone at roughly one-second latency and ran live in two San Diego clinics. Between them, he took engineering jobs at Deepgram, the speech-AI company, and at Whatnot, the live-shopping marketplace, before starting Sixtyfour and joining Y Combinator’s 2025 class.
Sixtyfour builds research agents that investigate individuals and companies for teams working in fraud, anti-money laundering, identity, and due diligence. In an independent evaluation by a prominent AI research lab, with results set to be published within the next week, Sixtyfour ranks as the most accurate system of its kind, ahead of tools from much larger labs. Shah’s standing rests less on the company’s size, which is modest, than on a position he argued early and consistently.
“Breadth is a vanity metric,” he said. “In fraud and identity, the only number that matters is how often you are right, and most AI research tools are wrong often enough to be unusable. One wrong answer and someone gets banned, fined, or even convicted.”
He first made the case to an engineer at a similar company, one that sells people data to sales teams. You can be wrong a third of the time in sales and still work off stale records, Shah told him, because the cost of a bad row is a wasted email. In fraud and identity cases, the same error rate results in someone being banned or arrested. It is an unfashionable thing to say in a market that has spent three years selling coverage, and his argument is that none of it counts if the answer is wrong when it matters.
What separates Shah from many founders selling AI is that he built the core of the product himself. He had chained language models with live web search into a recursive loop, one that finds a piece of evidence, uses it to look for more, and keeps going until it has the full picture, built specifically to investigate people. He knew it worked in January 2025, weeks before OpenAI released its own Deep Research, when the loop connected his own online profiles to a childhood pseudonym he had once used, a second digital identity most tools would never tie back to him. He architected that engine and the evaluation systems that measure its accuracy, the machinery that lets the company claim and defend its place on the public benchmark. In an industry where founders increasingly narrate technology someone else wrote, his authority is first-hand.
That conviction shapes what the company builds and whom it sells to. Sixtyfour’s customers cluster where a wrong answer is expensive, among blockchain-intelligence firms, payments companies, and the trust-and-safety teams at online platforms, rather than the sales departments most data vendors court. It is a smaller market than contact data in terms of the number of customers, but bigger in financial impact and more demanding.
He runs a team of twelve, eight of them engineers, and has hired beyond the usual founder network. Among his early recruits is a trust-and-safety leader with two decades in the field, formerly the head of trust and safety at Coupang, the South Korean e-commerce giant. It is a telling hire for a company that increasingly sells to the people who police online platforms.
Colleagues describe a founder unusually allergic to the unverified claim, applying to his own product the standard Sixtyfour sells its clients: a feature is judged less by how it looks in a demo than by whether its accuracy can be measured and defended.
Shah’s larger claim is that the discipline he works in is about to become everyone’s problem. As AI-generated fakes flood marketplaces and overwhelm verification systems, the platforms that run them are finding that good-enough accuracy is no longer good enough.
“Every marketplace is becoming a trust-and-safety company whether it wanted to or not,” Shah said. “One wrong call by our agents and someone gets punished, defrauded, or worse, arrested. The ones that survive will be the ones that can tell, quickly and correctly, who they are actually dealing with.”
The evidence that he was early is starting to accumulate. Marketplaces are reclassifying fraud as a trust problem that runs through the whole platform. Regulators are writing rules that assume real-time controls. The number of companies that need to know, quickly and correctly, who is on the other side of a transaction is growing faster than the tools that can tell them. Job marketplaces, rental platforms, and crypto exchanges that once outsourced trust to a vendor’s static file are now building verification into the core of their operations.
The recognition and the revenue have arrived together. The Aspire nod barely registered, he says: the goal is to keep building something people want and use. He has made a habit of arriving early to questions that later turn urgent, from emergency calls to clinical sensors to the identity of a stranger on the far side of a transaction. The field he has been quietly defining is starting to catch up to him.
