AI Product Engineer

3Y Health is building the operating system for healthcare practices, an AI-native platform that helps clinicians start, run and grow their practice. We work as a managed services organization (MSO): along with the software, our operations team runs the back office for the clinics we partner with, including billing, insurance and compliance. We currently serve applied behavior analysis (ABA) therapy practices and are expanding into speech, occupational and physical therapy, and psychiatry.

Members of Technical Staff own features end to end. You'll work with the clinicians and operators who use what we build, scope the problem, design and build the solution across the stack, deploy it and follow how it performs in production. Our engineering team is flat, with no sub-teams, so you'll work across the whole platform. We build with AI coding agents every day and expect you to do the same.

• Design and build AI agents and automated workflows for clinic operations, such as insurance benefits verification, prior authorizations, patient intake, claims and clinical documentation

• Build features for clinicians, patients and families, and our internal operations team, on web and mobile

• Take ambiguous problems from 0 to 1, working with operators and clinicians to define the problem, write the spec and ship a first version

• Design the services, data models and APIs behind your features and manage the infrastructure as code

• Own your services in production: deployments, monitoring and incident response

• Write evals for model-facing features and track the metrics of the workflows you ship

• Work with protected health information (PHI) safely, with least-privilege access, no PHI in logs and auditable changes

• Build and maintain the skills, plugins, MCP integrations and review bots our engineers use with Claude Code and Codex

• Backend: Python (our primary language for new services), Kotlin with Hasura (GraphQL) on our established platform, plus TypeScript (NestJS) and some Go

• Frontend and mobile: React, Next.js, React Native

• Data: PostgreSQL (including pgvector), Snowflake

• Workflows and contracts: Temporal, gRPC/Protobuf

• Infrastructure: AWS (ECS Fargate, RDS), Terraform, GitHub Actions, Datadog, LaunchDarkly

• AI: Claude and OpenAI models, Claude Code and Codex, with our own skills, plugins, MCP integrations and an AI code-review bot

• 5+ years of experience building and shipping production software

• Recent, hands-on experience using or building AI harnesses: the layer around a model or coding agent (project instructions, tools and MCP servers, skills, hooks, subagents, evals), not only calls to a model API. You can show what you set up or built.

• At least one personal project you built and published on the web, live and usable by anyone, ideally a SaaS with payments working (e.g. Stripe). Share the link when you apply.

• A track record of owning products or features end to end, from an unclear problem to a system people depend on

• Strong full-stack skills in Python and TypeScript, with experience across frontend, backend, databases and cloud infrastructure

• Comfort working in a large, established codebase

• Hands-on experience building with LLMs, including tool use, agent orchestration, structured outputs, retrieval and evals (LangGraph, an agents SDK or similar)

• Daily use of AI coding agents such as Claude Code, Codex or Cursor. You know how to set them up for a codebase (project instructions, skills, MCP servers, hooks, subagents), run several tasks in parallel and review their output carefully. If you haven't heard of Claude Code, this role is likely not a fit.

• Good product sense and the ability to make solid UI decisions without a dedicated designer, using our component library

• High agency: you unblock yourself, look for context and ask questions when you need to

• Clear written communication. The team is distributed across time zones and most decisions happen in specs, PRs and tickets.

• Experience in healthcare or another regulated industry (HIPAA, payer and claims workflows, EHR integrations, HL7/FHIR)

• Temporal or another durable workflow engine

• React Native

• Running LLM evaluations or safety checks on model output in production

• Having built your own agent harness, MCP server or plugin for a coding agent

• Open-source work or internal tools that other people use

If you meet all the must-haves and most of the rest, we'd like to hear from you, even if you don't check every other box.