Founding Product Engineer
ForminHQ is hiring its first engineer in New York City: someone with product sense and an entrepreneurial streak to help decide what Formin builds, then build it end to end.
About Formin
NYC property owners and managers track their buildings' violations across DOB, ECB, and OATH, separate city portals that don't talk to each other. Formin pulls all of it into one view and then helps resolve it. An AI agent reads each violation's regulatory context, drafts a resolution plan with task dependencies, and tracks the work through to closure. The product is live on real city data. Next is the regulatory intelligence underneath it: a proprietary NYC knowledge base, retrieval with real citations, agents that can look things up themselves, and evals that keep them honest.
The role
You'd be the first engineer, working directly with the founder on what gets built as well as how. You'll talk to property owners and managers, push back on the roadmap, and cut features that don't earn their place. Then you'll own the work from problem to shipped, from the database migration to the interface to the agent prompt. Some weeks that means a sync pipeline for a messy city dataset. Other weeks it means a retrieval layer, with evals to prove it beats the current answers.
Why Formin
- An unglamorous market with real money and bad software. Owners still rely on expediters, spreadsheets, and disconnected city portals, and nothing ties them together.
- A say in the product and the company, not just a slice of someone else's system. Pricing, design, and go-to-market are all on the table.
- A human-in-the-loop system, not a chatbot. The agent proposes, people review and decide, and the outcome depends on inspectors, hearings, and agencies nobody controls. Designing the work between the person and the AI when the result isn't deterministic is the core problem.
- Work that compounds. The regulatory corpus, the case history from real resolutions, and the structured knowledge behind them are assets a general-purpose model won't absorb.
What you'll do
- Own product areas end to end. Stay on a problem from the first unclear customer conversation through design, build, launch, and iteration.
- Do your own discovery. Talk to owners, managers, and expediters, watch how they actually work, and find the real problem behind the request. Push back when the ask doesn't match the problem.
- Build AI where the answer has to be right. Retrieval over NYC rules with citations, tool calls against DOB, ECB, OATH, and PLUTO data, structured extraction from summonses and city records, and evals that judge every prompt or model change against a golden set.
- Design the human in the loop. The agent proposes and people decide. Build the review, correction, and approval flows that make that work, and that let the system learn from every correction.
- Tame messy public data. Daily syncs from NYC Open Data, where the same lot number is padded two different ways and the same violation lives in two systems.
- Make and write down the technical calls. Choose the architecture, data models, and tradeoffs, and document them for the engineers who come after you.
- Ship in phases. Get working software to users early, keep it readable and tested, and iterate with customers instead of betting on one big launch. Weigh competing requests and make clear calls on impact, with little process.
What we look for
- Product intuition and a 0-to-1 track record. You think as hard about what to build as how to build it, and you've built something from nothing: a startup, a side project, or a new product inside a bigger company. Ideally you stuck around after launch and had to maintain it.
- Moves quickly. You don't need a spec to start. You also know where shortcuts come back to bite you, like sync jobs that quietly drop records or permissions that leak data.
- Close to customers. You'd rather talk to a customer than wait for a spec, and you can explain a tradeoff to anyone.
- Full-stack depth. You've shipped production software end to end, you're comfortable in Postgres, and you've built with LLMs past the demo stage, including structured output, tool calling, and evals.
- Detail oriented. In compliance, a wrong date or a missed hearing costs someone money. "The model said so" isn't good enough.
- Intellectually curious. A strange edge case in a city dataset is the best part of your day.
- AI native. You use AI across your whole workflow, can show how it makes you faster, and care more about the right solution than about being right.
Nice to have
- Experience with retrieval or search, such as embeddings, chunking, and hybrid ranking
- Experience designing human-in-the-loop systems, where AI drafts the work and people review, correct, and approve it
- Joined a company early and stayed while it grew
What we offer
Salary plus meaningful founding equity, and a real say in the product and the company from day one, not just the code.
Reach out if you're interested.
Send a link to something you've built, plus a few sentences on how you'd help us.
alex@forminhq.com