Anant
Bhardwaj

Bihar is not the obvious starting point for a Silicon Valley unicorn story. India's most populous state, landlocked and frequently cited in debates about development, is where Anant Bhardwaj grew up before he made his way to Pune for an engineering degree, then Stanford for a master's, then MIT's CSAIL for a PhD he never finished. The chain of moves was not strategic so much as relentless - each step toward a harder problem.

At Stanford, a professor named John Ousterhout handed him a piece of advice that stuck: use your intuition to ask questions, not answer them. Bhardwaj filed that away. It would take a few more years and a PhD database group at MIT before the question crystallized. His thesis research - a collaborative data platform called DataHub - kept bumping into the same wall. The systems his field studied were powerful but they all assumed your data was already clean, structured, labeled. In practice, enterprise data almost never is.

In July 2015, he left the PhD program, packed his work into a company called Instabase, and moved to San Francisco. The initial pitch was quiet - Greylock Partners and New Enterprise Associates put in $3.75 million without much fanfare. For two years after that, Instabase operated in something close to stealth. Four people. No public product. Just a slow accretion of enterprise customers who had discovered that their documents - the contracts, the invoices, the regulatory filings, the loan applications - were completely invisible to their software stacks.

The 2017 Series A - $23.2 million, led by Martin Casado at Andreessen Horowitz - gave Instabase its first serious fuel. Still, Bhardwaj resisted the pressure to announce, to hire aggressively, to build the sales org before the product was truly ready. A lesson he absorbed the hard way: he hired a sales leader too early, found the process wasn't scalable, and had to walk it back. The rule he took from that: validate with individual contributors first. Hire leaders only when the playbook is written.

By 2019, the playbook was written. Instabase came out of stealth on Bloomberg TV and simultaneously announced a $105 million Series B led by Index Ventures. Valuation crossed $1 billion. Four years after founding, Bhardwaj had a unicorn. He got there by spending those years doing something most startup founders cannot make themselves do: waiting until the thing actually worked.

The next evolution was not a pivot so much as a tectonic shift. When transformer-based language models arrived and then exploded into GPT-3, GPT-4, and the broader LLM wave, Bhardwaj had to make a genuinely painful call. Instabase had spent seven years building an operating-system-style architecture - data and execution layers cleanly separated, microservices all the way down. That architecture worked. It was elegant. And it had to go.

"We abandoned our 7-year OS architecture for LLM-as-OS," he said later, with the understatement of someone who lived through it. The company migrated its entire platform to transformer-based deep learning and launched pre-trained foundation models that customers could fine-tune on their own document types. The AI Hub followed in 2023 - a repository of AI applications built around generative AI tools, including a partnership with OpenAI that brought GPT-3.5 and GPT-4 into Instabase workflows.

The results are tangible. Instabase's customers now include four of the five largest U.S. banks, NatWest, Standard Chartered, AXA, Uber, and the U.S. Patent and Trademark Office. Annual revenue passed $50 million in 2024. In January 2025, the company raised $100 million in a Series D led by Qatar Investment Authority, with Andreessen Horowitz, Greylock, NEA, and Index Ventures reinvesting. Valuation came in at $1.24 billion - down from the $2 billion Series C mark.

That valuation decision says something about how Bhardwaj operates. In a market still obsessed with paper valuations, he accepted a down round with better terms rather than chase a headline number with strings attached. Goldman Sachs named him one of the most exceptional entrepreneurs of 2023. He took the award and went back to building.

Inside Instabase, the company practices what it sells. More than 80% of employees are weekly active users of the internal AI search product. The company has a standing OKR to increase productivity 50-75% using its own AI tools. Legal documents, HR queries, investor questions - all routed through the platform. Eating your own cooking, at scale, with measurable targets.

His favorite book is The Singularity Is Near by Ray Kurzweil. His advice to founders - delivered with a straight face - is simply: "Don't follow any advice." He still shows up as "on leave" on his old MIT CSAIL page, technically mid-PhD after a decade of building a unicorn. The question his dissertation was circling around - how do you build systems that work with data as it actually exists, not as it should be structured? - Instabase is one long, generative answer.