🧠 About Legion Health

Legion Health (backed by Y Combinator) is building autonomous mental healthcare—AI that takes real responsibility across the full care loop: before the visit, during the visit, and after the visit.

We run our own psychiatric practice, which means we ship into reality—not theory. We’re rebuilding the patient experience end-to-end:

• Outside the visit: scheduling, intake, benefits checks, paperwork, reminders, care coordination, billing, risk detection, follow-ups, and patient messaging

• Inside the visit: visit support, real-time note scaffolding, structured history capture, differential reasoning support, diagnostic signal extraction, treatment planning assistance, medication decision support, and clinician workflows that make great care feel effortless

Not “doctor-in-a-box.” Not a layer of prompts. We’re building AI teammates that clinicians can trust—agents that do the work, show their reasoning, handle edge cases, and escalate cleanly when humans need to step in.

Our agent system already supports 2,200+ patients with a tiny team. The systems you build go straight into production and show up in patient outcomes, clinician velocity, and the quality of care—immediately.

Legion exists to make mental healthcare fast, affordable, and excellent at scale. This role is one of the highest-leverage ways to get there: you’ll help turn clinical judgment into reliable systems, and turn AI from a tool into a true operator—so patients get sharper care and clinicians get their time back.

🚀 Role: Founding Engineer – AI-Native Mental Health Care (In-Person in SF)

Hey—I’m Daniel Wilson, co-founder & CTO at Legion. I studied applied math + CS at Princeton, then led the Office Scripts + Power Automate integration at Microsoft—shipping the highest retention-driving feature in the product. Before that, I did deep learning + RL research. I’m obsessed with building AI systems that make real workflows cleaner, safer, and more autonomous. Github: https://github.com/danielgwilson

I’m hiring a Founding Engineer to own core product + agent systems end-to-end. You’ll build the layer that coordinates patients, clinicians, and AI agents as one coherent machine—tight feedback loops, high standards, and real ownership.

This isn’t “just infra.” It’s building the product heart of an autonomous care system: the workflows, the agent behaviors, the guardrails, the integrations, the UX that makes it usable, and the reliability that makes it trusted.

Roles evolve quickly here—including mine. I’m looking for someone who wants to carry real weight, take full ownership, and grow into running major parts of the system—not someone who needs a perfectly scoped box to operate inside.

✅ What You’ll Do

• Own our event-driven backend – Architect and scale our Node.js / TypeScript / Supabase (Postgres) / AWS stack. Design schemas, invariants, and workflows that encode how psychiatric care actually operates. Turn messy real-world processes into clean state machines and event streams.

• Build real LLM agents as coworkers – Implement tool use, retries, memory, and safety rails. Design action schemas and evaluation loops so agents can run reliably in production. Work on orchestration, context management, and multi-step workflows.

• Shape human + AI ops UX – Build internal tools used by both humans and agents. Make it trivial to see “what happened, why, and what should happen next” in any patient journey.

• Define world-state & simulation – Model the canonical state of a patient’s journey across time. Power alerting, routing, and decision-making from that live simulation.

• Own data, safety & compliance – Engineer HIPAA-compliant pipelines for transcripts, events, and EHR data. Ensure PHI access, agent actions, and human overrides are all auditable.

• Drive architecture & strategy – Work directly with me to debate tradeoffs, define new primitives, and decide what we build next.

You don’t need to be a pure backend engineer—we value great product engineers who can learn fast, make strong simplifying decisions, and grow into owning backen