Most people finish a PhD and take a nap. Zohar Bronfman finished two of them and started a company two months later.
Today he runs Pecan AI, a predictive analytics company that lets business teams build machine-learning models without hiring a single data scientist. The customer list runs through Fortune 500 and Fortune 100 companies. The funding runs to roughly $127 million, led by Insight Partners with checks from GV - Google's venture arm - and Dell Technologies Capital. The product now spans everything from customer churn to demand forecasting to a predictive modeling co-pilot the company shipped in 2025.
But the version of Bronfman worth knowing isn't the funding total. It's the man who looks at a room full of AI hype and says, plainly, that the emperor is underdressed. "I don't think LLMs are taking us anywhere closer to AGI," he has said - an unusual sentence from someone whose company sells AI, and a very usual one from someone who spent years in a neuroscience lab studying how prediction actually works inside a skull.
When used alone, ChatGPT and other similar tools based on large language models can't solve predictive business needs.Zohar Bronfman, on the limits of the tools everyone else is selling
A company born from a deadline they missed
Here is the strange, specific beginning. Bronfman and his co-founder, Noam Brezis, met as master's students in cognitive science at Tel Aviv University. The two of them entered an international data science competition and built a piece of data-preparation automation for it. They missed the submission deadline. The entry never counted.
And that lost entry became the prototype for Pecan AI. Instead of shelving it, they kept building. Two months after both of them wrapped up their doctorates in 2018, they rented a small room at the same university they'd just graduated from and started pitching venture capitalists with, by their own account, very little business experience. It worked. Haim Sadger and Aya Peterburg of S Capital wrote the first $4 million check.
There is a lesson buried in that origin. The deadline you miss is sometimes the one that frees you. A competition entry has to fit the rules. A company doesn't.
Two doctorates, no lane
Bronfman didn't pick a specialty and burrow. He earned a PhD in computational cognitive neuroscience and, at the same time, a PhD in the history and philosophy of science and technology. Underneath both sits a BA in economics from the Open University of Israel. The combination sounds like a résumé that couldn't decide - until you watch how he uses it.
The neuroscience tells him what prediction is: a living system guessing the next moment from messy, incomplete signals. The philosophy of science tells him to keep asking whether a tool actually does what its makers claim, or merely looks like it does. Put those two habits together and you get a founder unusually allergic to hype, unusually comfortable saying a popular technology is oversold.
The honesty problem
One of Bronfman's sharper observations is also his most uncomfortable one for the industry he works in. "You can sell things, especially to larger organizations, that actually don't have ROI," he has said. Big companies buy software that looks impressive and quietly changes nothing.
He built Pecan to be the opposite bet, and he enforces it in ways that cost money. He has said he's willing to eliminate features that turn a profit if they don't help customers stay for the long run. That is not the standard growth-at-all-costs playbook. It's the position of someone who'd rather answer one honest question - will this model actually move your revenue? - than win a conference demo.
Our mission has always been to democratize data science, and we've identified a simple way to marry the power of predictive analytics and Gen AI.On launching Predictive GenAI, 2024
Marrying the two halves of AI
In January 2024, Pecan launched what it calls Predictive GenAI - a fusion of the predictive modeling the company was built on with the generative AI everyone else was chasing. Bronfman's framing is that neither half is enough alone. Generative AI can explain and interface; predictive AI can actually forecast an outcome. Bolt them together and a business person can describe a problem in plain language and get a working predictive model on the other side.
The through-line across every Pecan release is the same target he picked at the start: the mid-market. Companies too small to keep a data science team on payroll, too serious to run their future on a spreadsheet. Bronfman never tried to build universal AI. He tried to solve one problem - predictive modeling for teams without specialists - exceptionally well. That focus is also why he prefers small teams to big ones. Headcount, in his view, is not a scoreboard.
Where he sits now
Bronfman splits his orbit between New York, where the company is increasingly centered, and Ramat Gan, where Pecan was born in the heart of Israel's startup scene. He sits on the Forbes Technology Council, turns up regularly on data-science and AI podcasts, and remains one of the more quotable skeptics in a field that rewards evangelists. His measure of a good day is refreshingly unglamorous: "A good day is a day I am closer to accomplishing my goals."
It's a tidy summary of the whole enterprise. No fireworks. Just a scientist, two doctorates deep, quietly building a machine that guesses what happens next - and refusing to pretend it can do more than it does.