Founding Engineer

About Hedge Hedge is a specialty insurance brokerage built from the beginning to be operated by AI. Software agents run the core of the business in production today: they read broker submissions, select markets, complete carrier applications, negotiate quotes, prepare binders, issue documents, and handle regulatory filings. A small team directs and supervises that work through an internal command system rather than doing it by hand. Many companies describe themselves this way. Very few are structured so that it has to be true. We are, and it shapes every engineering decision we make. The problem space Commercial insurance is one of the largest industries in the world, with more than $1 trillion spent in 2025. It still runs on email, PDFs, phone calls, and carrier portals from another era. It is also genuinely difficult terrain: fifty states of regulation, surplus lines taxation, shifting carrier appetite, documents that carry legal weight, and real money moving between counterparties. That difficulty is the opportunity. This industry has resisted automation precisely because it is nuanced, relational, and unforgiving of error. Automating it well requires unusual engineering, and the companies that manage it will be very large. The work You would be our founding engineer, building across the full surface of the company: Agent infrastructure. Long-running, multi-step, interruptible agents operating across email, documents, APIs, and live browser sessions, held to the reliability standard of a business whose output is binding legal paper and moving money. The command system. The internal console through which a small team supervises a fleet of agents. Closer to an operations center than an admin panel. Customer surfaces. A broker portal, public API, MCP server, and CLI. Integrations. Carrier APIs where they exist; portals, email, and document parsing where they don't. E-signature, payments, state filing systems, and the long tail of legacy software this industry runs on. In time, our own models. Insurance-specific evals, fine-tunes, and domain models built on data and feedback loops that exist nowhere else. When part of a workflow resists automation, we treat that as an engineering problem to solve, not a place to quietly add headcount. Holding that line is the hard part of the job, and the interesting part. Stack: Python (FastAPI, SQLAlchemy, Celery), TypeScript and React, Postgres, Redis, LLMs across providers, browser automation. The way we build mirrors the way we operate: agents carry a large share of the engineering itself, and we adopt, extend, and discard tooling quickly as the frontier moves. You will have a strong hand in where all of it goes. Who we're looking for Someone whose instincts run closer to a founder's than to an employee's. In practice, four things: Engineering depth. Three or more years building production systems end to end, with real LLM experience beyond thin wrappers: agents, tool use, evaluation, document pipelines. You have shipped things people depend on. An AI-native way of working. You already build with frontier models and agentic tooling as a matter of course, and you rework your own setup as the frontier moves rather than waiting for practices to settle. How a company works internally tends to become its product; we hold our own workflows to the same standard as the systems we ship, and we expect you to push that envelope, not just keep up with it. Appetite for the work. Companies at this stage are built on persistence as much as insight. We are direct about this: the pace is demanding, and the role suits people who do their best work with a lot on the line. Commercial judgment. You understand how businesses make money and let that shape what you build. Between the elegant abstraction and the thing that wins a customer this week, you know which matters now. If you have built or run something of your own, you have likely learned this firsthand; we weigh that heavily. What the role involves Engineering here is not insulated from the business. You will talk with the brokers and carriers who use what you build. You will spend time inside live operations, supervising agents on real business, because that is the only way to understand the domain well enough to automate it. You will be measured on outcomes rather than output: policies bound, customers kept, revenue earned. And you will hold meaningful equity, with everything that implies about taking on the problems that are nobody's job. Details \- Work directly with the founders and set the technical culture from day one \- Real revenue, real customers, agents in production \- $150k to $300k, 1% to 3% equity, San Francisco