Staff Software Engineer, Frontend

We're building the AI agents that run insurance — a trillion-dollar industry still powered by manual work. We've raised $30M+ from a16z, YC, Nexus, South Park Commons, and Converge, grown 10x in the last 12 months, and work with top 10 insurance carriers and brokers, plus Fortune 100 companies. Both founders are engineers, and engineers have an outsized say on product direction.

Nobody owns our frontend today. It's grown to ~690K lines of TypeScript/React across ~2,250 files — a workflow builder, a citation-backed document viewer, data grids holding tens of thousands of rows, a chat assistant, and live views of agents running for minutes at a time. Features ship every week, but the parts nobody owns pile up: two generations of the workflow builder running side by side mid-migration, the same component built three times, performance with good tooling and no one driving it.

We're not hiring you to clean up after us. We're hiring you to decide what "good" looks like here, say it out loud, and make it true — while the product keeps shipping.

• Architecture and stack. State, data fetching, routing, module boundaries, build tooling — what we keep, what we drop, what we add. Today: React, TypeScript, Vite, Tailwind, Radix, TanStack Query, Bun. You decide what it looks like in a year.

• One component system. We have ~90 shared primitives and 231 Storybook stories, and we still write the same component twice. Pick the one that wins, migrate the product onto it, and put a gate in place so it stays that way.

• Performance as a number, not a vibe. Our home route ships ~567KB of gzipped JS before a user can interact. We have bundle budgets, Lighthouse in CI, real-user vitals, and a Sentry digest — nobody's driving them. Own the number, move it, and tell us by how much.

• Product craft. Work with our designers as a peer — critique their work, fill the gaps, hold opinions on layout, density, motion, and hierarchy, and argue for them.

• A codebase that's safe for coding agents to touch. A large share of our code now arrives from coding agents. Typed boundaries, lint rules, visual regression tests, and fast feedback loops are what make that safe at our speed — you design for it.

• The engineers around you. Review, mentor, and hold the bar on every frontend change, across India and the US.

Two concrete examples of the kind of problem this role exists to solve: converging our two live generations of the workflow builder into one without freezing the product for a quarter, and designing the interaction patterns for agents that stream partial results over minutes — there's no established playbook for this yet, you'd be inventing ours.

• 8+ years building production frontend, deep in React and TypeScript — you've shipped customer-facing product fast and lived with the consequences afterward.

• You've owned a frontend codebase, not just features inside one: architecture, build tooling, component systems, test strategy, release safety.

• Demonstrated performance practice — you've held a number and moved it. We'll ask which one and by how much.

• Real product taste — you can look at a screen and say precisely what's wrong with it and why.

• Sharp with coding agents — you push back on what they write rather than merge it on faith, and you build the checks that make agent-written code safe to ship.

• A clear writer — we make most decisions in writing, across time zones.

• Bonus: design systems experience, data-dense or document-heavy products, 0-to-1/early-stage work, B2B SaaS, LLM-backed interfaces.

Remote in India, Bangalore preferred (where most of engineering sits), but open countrywide. Our designers and part of engineering are in San Francisco; you'll overlap with them roughly 9 PM–2 AM IST. Outside of that window, working hours are flexible to manage.