In January 2024 Bhavish Aggarwal made India's first AI unicorn. Eighteen months later, he was cutting staff.
He was born in Ludhiana in 1985 and studied computer engineering at IIT Bombay, graduating in 2008. He then worked at Microsoft Research India for about two years, filing two patents and publishing three papers. A bad taxi ride is the story told about why he left to start Ola. That is how a researcher became a mobility founder.
He then worked at Microsoft Research India for about two years, filing two patents and publishing three papers.
The personal record on Krutrim is thin. We know the founder's money, not his doubts. TechCrunch reported that Aggarwal put in $230 million of his own, a figure we could only confirm through a search summary. A $50 million round led by Matrix Partners India in January 2024 valued Krutrim at $1 billion.
The idea was to build India's own large language model, trained for Indian languages, with Indian chips and cloud behind it. For a moment it looked like the next Ola: loud, fast, backed by its founder's conviction.
The story after that is about the gap between a promise and a product. In 2025, after the Kruti assistant launched, about 100 employees were let go. Reports say headcount fell from more than 550 in August 2025 to roughly 150 or 160 by March 2026, and that Kruti was pulled from app stores in April. A later round in 2026 reportedly cut about half the remaining staff. Fundraising targets fell, one report says from $500 million to $300 million, because investors showed little interest.
The company's answer was to change the shape of the business. LLM and chip work paused, and AI cloud services came to the front. In May, Krutrim said revenue for fiscal 2026 had crossed Rs 300 crore and that it was profitable. We could not find independent confirmation.
The lesson for first-time founders is hard and specific. A valuation is a promise, not a market. Aggarwal's backing bought time to find out which parts of the AI stack Krutrim could actually sell. The founder who can afford to be wrong for longer is not safe from being wrong. He is just wrong at a larger scale.