Founding Engineer (AI Agents)

About the job
The Opportunity
A stealth AI startup in San Francisco, founded by Harvard and MIT alumni, is building agents that handle complex, high-stakes paperwork for US families. They research what applies to each family, handle correspondence, and follow up over weeks. The product is running in test. You’ll be the first engineering hire, working directly with the two co-founders to take it live.
You’re a Fit If YouAre high agency. You don’t wait for permission or perfect information. You see what needs doing and make it happen.Figure things out on your own. When something’s unclear, you research, decide, build, and check in with what you did, not with a list of questions.Are comfortable in an unstructured environment: no detailed specs, no fixed processes, no one assigning your dayHave owned a product or major feature end to end, ideally as an early engineer at a startupShip fast, with small changes daily, and know when scrappy is good enoughAdapt easily when priorities shift mid-week. The product is early, and plans will change as users teach us things.Make product calls yourself and push back when you disagree, including with the foundersKeep the team updated on what shipped, what’s blocked, and what needs a decision
What You’ll Build
Features end to end, from agent logic to the screen a family usesAgents with tool calling, guardrails, evals and tracingLong-running workflows that wait days for replies and survive restartsSecure handling of sensitive personal data
Must-Haves: 5+ years of software engineering, with strong TypeScript across the stack (Node.js + React)Independent production ownership: you’ve designed, deployed and run features in production yourself, including monitoring, debugging live issues and fixing what breaks, with no separate DevOps or QA team to lean onBuilt and shipped at least one LLM-powered feature or agent yourself: tool calling, structured/validated output, and handling model failuresPostgreSQL and relational data modellingBackground jobs, queues or workflow engines (e.g. Temporal, Inngest, Celery) that retry and survive failureDeploying on AWS or Azure: CI/CD, environment setup, logging and alertingSecure handling of sensitive data: access control, encryption, no secrets in code or logsAutomated testing as a habit, including tests for model-driven behaviourDaily use of AI coding tools, with the judgement to review what they produceExcellent English; comfortable on camera for daily stand-ups
Bonus: founding-engineer experience; Python, Go or Rust in production; agent frameworks (LangGraph, Claude Agent SDK, OpenAI Agents SDK); evals and observability for LLM systems; infrastructure as code (Terraform/Pulumi); document processing and OCR.
DetailsLong-term engagement through Metric Tree Labs, dedicated to this clientIST hours: 2:30–11:30 p.m. or 7:00 a.m.–4:00 p.m. (finalised after the first few weeks)Own laptop (16 GB+ RAM, disk encryption on)Process: Technical task, founder intro call → 90-min technical session. Quick decisions.Start: Immediately on finalisation
To Apply Send your CV, a link to something you’ve shipped (repo, live product or demo), and 2–3 lines on a time you solved a problem with no guidance and no one to ask: what you did, and how fast you got it live.Requirements added by the job poster5+ years of Software Development experience

Skills
Python, JavaScript, TypeScript, React, Node.js, Golang, Rust

Experience
5+ yrs

Work arrangement
Yes