Founding Engineer

THE ROLE: Founding Engineer. Full-stack, leaning hard into ML. You'd own the product surface and the agent layer underneath it, with real authority over the architecture and where it goes next. WHY WE NEED YOUYou've probably been the person on your team who actually understands how the system behaves in production — not just how it was designed. You've shipped ML into something real, watched it degrade in ways the offline metrics didn't predict, and built the tooling to catch that next time. You're tired of having the interesting architectural decisions made two levels above you. You'd rather own a hard problem end to end, with the authority to decide how it gets solved, than optimise a slice of someone else's. THE HARD PROBLEMSAgents that run for months, not minutes. Durable state, replanning, idempotency, recovery. No framework does this. Credit assignment on sparse, delayed, confounded outcomes — and the offline eval to ship against it, because you can't A/B against real money. Autonomy with hard guarantees. Actions hit a real ledger. "Probably right" isn't available. Plus the usual, done properly: multi-tenant isolation, high-throughput reconciliation, exactly-once scheduling, flaky third-party APIs, inference cost and latency, observability that still answers questions six weeks later. ALSO YOURSModel strategy — what runs where, and the quality/cost/latency tradeoff per decision type. Deep integration with systems of record — not just reading, but acting in them safely and reversibly, to a standard an auditor accepts. The product itself — onboarding, dashboards, the control plane that makes a team willing to switch autonomy on. Plus the founding-engineer tax: CI, observability, on-call, the security questionnaire that unblocks a deal.

WHAT WE'RE LOOKING FOR: Strong Java and Python, plus real ML engineering experience. Backend discipline alongside the ability to train, evaluate and ship models that survive production. No fixed years requirement — but we'll want referenceable experience, not a reading list. You've shipped production systems and owned them afterwards. ML and LLM work beyond prompting: fine-tuning, evals, retrieval, agent orchestration, or the cost/latency tradeoffs that decide what's viable. Comfortable owning product surface as well as models, and comfortable being the one who decides. Helpful, not required: fintech, accounting or ERP domain experience; a prior founding-engineer role; financial-grade correctness and audit requirements; anything public that shows you ship. THE STACKJava, Python and TypeScript. Anthropic API alongside open-weight models. Digital Ocean, GitHub, Google Workspace. Integrations into NetSuite, Business Central, Sage Intacct and QuickBooks. Slack is a first-class surface, not a notification channel. You'd have real influence over what this becomes — several of these decisions are still open.

HIRING PROCESS: Two steps. No take-home marathon, no LeetCode. Intro call with the founder. What you've built, what you want to own, how the comp structure lands for you. Solve a real problem — one session, a problem from our actual domain. Use whatever you like, including Claude Code, Cursor or any other agent. We care how you think and how you use the tools. COMPENSATIONFounder-level equity, 0.5–5%, vesting on a standard schedule from your start date. No cash salary until our seed round closes, at which point the role converts to salary.

HOW TO APPLY: Apply here, or email with something you built that you're proud of and what was hard about it, a repo or demo, and what you'd want to own first. No cover letter. We reply within 72 hours either way.