CyrusKhajvandi

A reading problem he refused to live with

There is a particular kind of guilt that haunts scientists: the stack of papers you mean to read and never will. Cyrus Khajvandi knew it well. A Stanford biology graduate who had spent years at the lab bench, he kept falling behind on the research that mattered to him while the daily grind of building companies ate his hours.

Most people accept that backlog as the cost of a busy life. Khajvandi treated it as a bug. He noticed the problem wasn't his alone - grad students, postdocs, even seasoned principal investigators all drowned in the same flood of dense documents. And by 2022, the tools to do something about it had finally arrived. So he started building an AI that could read what he didn't have time to, summarize it, and answer questions about it.

That project became Humata. The pitch is disarmingly plain in his own words: "Humata is like ChatGPT for all your files. You can ask Humata any question about your PDF and automatically get the answer with highlighted references, so you can fact check it and guarantee correctness." The phrase "fact check it" is the tell. Khajvandi came from a world where a wrong citation can sink a paper, and he refused to ship an oracle you had to trust blindly.

An answer is only as good as its receipt

Upload a document. Ask a question. Get an answer that points straight back to the line in the original file where it lives. That side-by-side highlight is the whole philosophy: Humata only works from the data you give it, and it shows you exactly where every claim came from. No hallucinated confidence, no unsourced summary you have to take on faith.

It turned out the appetite for that reached far past academia. The early crowd was researchers and students, but lawyers wanted it for contracts, oil and gas teams wanted it for technical filings, customer-support crews wanted it for knowledge bases. The platform launched in February 2023 and, within months, was chewing through tens of millions of pages for a user base in the millions.

Google and Cathie Wood walk into a seed round

In October 2023, Humata closed $3.5M led by Google's Gradient Ventures, with Cathie Wood's ARK Invest and M13 joining in. It is a strange and telling cap table: the search giant's AI fund and one of the most-watched public-market investors both putting early money behind a document-reading startup out of Austin. The company also began pursuing SOC 2 certification, the unglamorous paperwork that opens enterprise doors.

His read on what the money buys is bigger than features: "The long-term vision for Humata is to develop tools that make people wiser and more productive. The future of work will rapidly change this decade as AI automates mundane processes and frees people to be more creative."

A founder who keeps starting over

Humata is not Khajvandi's first rodeo, or his second. He has a habit of building - across biotech, fintech, and security - and a record of seeing things through. The throughline isn't an industry. It's a pattern: spot a painful, unglamorous problem, then go build the thing that removes it.

The chess champion in the other chair

Every founder story has a co-founder it can't do without. Khajvandi's is Dan Rasmuson, Humata's CTO, who had already co-founded Labelbox on its way to a $1B valuation. He is also a Forbes 30 Under 30 honoree and a national chess champion - the kind of partner who tends to see a few moves ahead. Rasmuson joined as the platform launched, and the two have run it together since.