Zubin Koticha

Catching the failures your AI will never confess to

An AI agent rarely tells you when it has gone wrong. It just keeps answering, confidently, while something underneath quietly rots. Zubin Koticha built a company around that single uncomfortable fact.

That company is Raindrop, and the shorthand he uses for it is blunt: Sentry for AI. The pitch is that AI engineers spend their days flying blind, shipping agents into production and finding out about the broken behavior days later, usually from an angry user rather than a dashboard. Raindrop watches the conversations and execution traces, then sends an alert the moment an agent misbehaves, with a link straight to the event that caused it. Find it, track it, fix it. The unglamorous plumbing of trust.

In December 2025 the company announced a $15 million seed round led by Lightspeed Venture Partners, framed around a deceptively simple promise: detect the critical failures in AI agents before they cost you. Raindrop had already passed through Y Combinator's Winter 2024 batch, and along the way shipped Deep Search, which reads across millions of production events the way Deep Research reads the web, surfacing the patterns a human would never have the patience to find.

Koticha started it in 2023 with two people he trusts completely: Alexis Gauba, who he had already built a company with, and Ben Hylak, who spent four years on Apple's Human Interface team shaping visionOS. It is a strange, complementary trio. A derivatives obsessive, a financial-instrument inventor, and an interface designer from the robotics-and-avionics world, all pointed at the same problem of making machines legible.

A New Yorker who learned to code on the other coast

He grew up in New York and went west to UC Berkeley, where two things happened that set the rest of the story in motion. He taught himself to code, and he fell in with the crowd around Blockchain at Berkeley, the campus group that produced an unusual number of crypto founders in the late 2010s. That is also where he met the cofounders who would follow him from one venture to the next.

Berkeley gave him the obsession he still nurses for hobby: esoteric financial derivatives. Most people find options pricing a chore. Koticha treats it as a puzzle worth chasing for its own sake, which turns out to be the rare hobby that doubles as a business model.

Opyn, Squeeth, and a sale to Coinbase

Before AI agents, Koticha was deep in decentralized finance. He cofounded Opyn with Alexis Gauba and built it into the first and largest DeFi options platform, eventually crossing $15 billion in volume. The investor roll behind it read like a who's-who of the era: Paradigm, Balaji Srinivasan, and Figma's Dylan Field among them.

Opyn's signature invention was Squeeth, short for squared ETH, the first of a new asset class called power perpetuals. The idea was to give traders continuous, options-like exposure to ethereum squared with no strikes and no expiration dates to manage. The research came out of a team that included Gauba and Aparna Krishnan working alongside Paradigm's Dave White and Dan Robinson. It was the kind of thing that sounds like a math joke until it starts trading, and then it traded hard. Opyn was eventually acquired by Coinbase.

The throughline from Opyn to Raindrop is not crypto. It is the instinct to build the instrument panel for a frontier where nobody can yet see what is happening. First it was a new kind of derivative. Now it is a new kind of software that talks back.

Twice now, the same move

Look at the two companies side by side and a habit emerges. Opyn measured and packaged something the market could not previously hold in its hands. Raindrop measures something AI teams cannot otherwise see. In both cases the bet is the same: the boring layer that surfaces reality is more valuable than the flashy thing on top of it.

It is a contrarian's bet, and Koticha wears the contrarian streak lightly. His handle on X is @snarkyzk, which tells you he is not taking himself too seriously even while raising eight figures to babysit other people's robots. He writes the occasional piece for TechCrunch, supports Arsenal with the particular suffering that requires, dances bachata, deadlifts, and collects niche geography facts the way other founders collect grudges.

The bigger ambition sits underneath all of it. Software errors got Sentry. Web analytics got their dashboards. AI agents, the fastest-moving and least predictable software anyone has shipped at scale, still mostly run in the dark. Koticha wants Raindrop to be the default light switch, the thing every team flips on before they trust an agent in front of customers. If he is right, the most important product he ever builds will be the one users never see, working quietly in the background, catching the failures before they ever reach you.