Shaurya Aggarwal built chips at Google. Now he runs Florin (YC S26), an AI insurance carrier trying to cover the robots, rockets and reactors that legacy insurers won't touch.
Most people who spend their early career building the physical guts of artificial intelligence - the accelerator chips that make giant models run - do not wake up one morning and decide the most interesting problem left in the world is insurance. Shaurya Aggarwal did exactly that. He runs Florin, a company in Y Combinator's Summer 2026 batch, and its promise fits on a bumper sticker: insurance, bound in minutes.
Behind that short line sits a bigger idea. The industries that are supposed to define the next few decades - robotics, space, advanced energy, defense - are notoriously hard to insure. The risk is new, the data is thin, and the humans who price these policies are working from playbooks written for a world of warehouses and delivery vans. Florin's answer is to hand the job to software.
Aggarwal describes the core of the company as a world model: a system that ingests ACORD insurance forms, satellite imagery, public records and other signals, then runs tens of thousands of simulations to figure out what a policy is actually worth. The output is a bindable quote in under a minute, where the traditional process can take weeks.
Aggarwal's path to insurance runs through some of the least insurance-like work imaginable. At Google, he was on the next-generation TPU team, helping design the custom silicon that powers modern machine learning. Before that, he built greenfield encryption products - protecting data both at rest and in transit - that were deployed across more than 100,000 hosts. He holds a BS in Computer Science from Purdue.
That background matters, because it explains the shape of the bet. Someone who has shipped security at the scale of six-figure host counts, and who has worked on the chips underneath large models, tends to see the world as a set of systems that can be measured, modeled and automated. Insurance underwriting - slow, paper-heavy, human-gated - looks less like a sacred craft and more like a system waiting for better tooling.
Florin did not start life as an insurer. The earliest version pitched a financial operating system - banking, payments, cards and treasury folded into one account for startups, so founders could hold, move and grow capital without stitching together half a dozen tools. The ambition then was to own the money-movement rails end to end, on a single ledger.
Somewhere in the run-up to and through YC, that idea gave way to a sharper one. The company kept the name and turned toward insurance, a market that is enormous, unloved, and unusually friendly to anyone who can price risk better than the incumbents. It is the kind of change of direction that looks reckless from the outside and obvious in hindsight if it works.
Fig 1 - The underwriting loop Florin is trying to compress from weeks to seconds.
The pitch is that a full-stack AI carrier - one that owns the underwriting engine rather than reselling someone else's - can win on three fronts at once: loss ratio, speed and operating cost. Get all three right and the economics of insurance start to look very different, because the whole business is a wager on whether you priced risk correctly.
The frontier framing is doing real work here too. If you are a robotics company, a launch startup or a next-generation reactor project, you are asking a legacy carrier to underwrite something it has never seen before. That is precisely the gap an AI-native underwriter is built to close, by learning from data instead of leaning on decades of actuarial precedent that does not exist yet.
Fig 2 - The competitive story Florin tells, illustrative not audited.
Aggarwal did not go it alone. Florin's co-founders bring a mix of hardware, autonomy and scale. Amol Pant, the CTO, has worked on perception at a robotics startup, on Tesla's Autopilot computing infrastructure, on NASA Artemis rover prototypes and Boeing flight simulators. Aydin Sorensen spent years as a tech lead across AWS and Twitch and was a founding employee at a construction robotics company, where he helped scale revenue dramatically.
It is a deliberately physical set of backgrounds for an insurance startup - people who have actually built robots, chips and spacecraft-adjacent systems. If the plan is to insure the machines of the future, it helps to have founders who understand how those machines break.
BS in Computer Science, Purdue University.
Built encryption products deployed across 100,000+ hosts, then joined the next-generation TPU team.
Founded Florin with Amol Pant and Aydin Sorensen; joined Y Combinator's S26 batch.
Reported $130K seed round; Florin sharpens its focus into an AI-powered insurance carrier.
There is a quiet contrarian streak in choosing insurance. It is not the flashy corner of AI. It does not make for viral demos. But it is one of the largest markets on earth, and it rewards exactly the thing a chip-and-security engineer is trained to obsess over: getting the numbers right, at scale, without drama. Aggarwal is wagering that a small team with a good world model can out-underwrite institutions that have been doing this for a century.
Whether Florin becomes the default carrier for robots and rockets or a smart idea that ran into the hard realities of regulated insurance is still an open question. What is clear is the instinct behind it - point serious engineering at an industry that has been coasting, and see how much of the old process turns out to be optional.
For now, the company is four people, a seed check, and a claim that fits in five words. The rest is a matter of whether the simulations are right.