AI & Platform Engineer: Digital Mental Health

Why this role exists
Rethink Wellbeing delivers peer-facilitated group CBT programs that measurably improve mental health at a fraction of therapy's cost. Our AI+ CBT Lab program has shown to be likely feasible in our pilot project with the Karlsruhe Institute of Technology. We are testing a bold hypothesis: that careful, human-in-the-loop AI can cut the cost and expert hours of effective peer-delivered group therapy by an order of magnitude, without cutting corners on safety.
Three AI MVPs are already live: a video assessment AI to evaluate facilitators' behavior and performance, two coach chatbots, one for participants and one for facilitators. Also, the platform concept, demo, and feature spec are prepared! Your job: take them from "promising prototype the team babysits" to "production systems that know when to act and when to call a human."
What you'll build
Chatbot triage. Our Slack chatbots (built on Cassidy) support participants and facilitators with logistics questions and CBT/facilitation coaching. Today, humans monitor everything. You'll build the triage layer: the bot handles routine cases, while new, risky, uncertain, or complex ones escalate to humans. Reliably, every time.
Facilitator-quality AI. An AI that assesses facilitator behaviors (shown / not shown) and quality on a 0-10 scale from transcripts, audio, and video of role-plays and real sessions. The MVP is being validated for accuracy; you'll turn it into a human-in-the-loop pipeline that acts when confident and escalates when uncertain.
The platform. Group matching, role-play scheduling, deposits and refunds, and triggered communications. Rule-based first, AI-assisted next. Interactive demo: https://rw-cbt-lab-pilot.netlify.app/ · Full feature spec available on request.
What you'll own
You are both product owner and builder for the platform and its AI features:
Product: turn user needs into precise specs; decide rules vs. AI per feature; define evals, metrics, and monitoring; test before release; keep GDPR/privacy baked into every spec.
Engineering: full-stack development (Next.js/TypeScript, Supabase/PostgreSQL, FastAPI), LLM integration (Anthropic API, RAG), CI/CD, security for sensitive health-adjacent data, and monitoring. A live service carrying crisis flags must never fail silently.
Strategy is set by our Director. Clinical content and safety criteria come from our clinical psychologist, the clinical platform owner and your closest collaborator. You'll also work with our ops assistant, who runs UX tests, and receive senior technical supervision from our KIT research partners.
What we offer
Ship AI that's actually used: in live clinical programs, this quarter, not someday.
A rare combination of startup pace and research rigor (your systems feed peer-reviewed publications).
High autonomy, low bureaucracy, fully remote, flexible hours with EU overlap.
A mission with a measurable return: wellbeing gained per dollar spent.
What you bring
Full-stack web development you can prove (ideally React/Next.js, TypeScript, PostgreSQL/Supabase).
Hands-on LLM engineering: model APIs, prompting, RAG, and, critically, evaluating outputs against human ground truth.
Product judgment: you can talk to users, write a spec, and say "no" to scope creep.
The temperament for safety-critical work: you'd rather escalate a false alarm than miss a real one.
Fluent written English; comfortable async and remote.
Bonus: experience with health/clinical products or GDPR-sensitive data; Slack bots or Cassidy-type builders; audio/video ML pipelines; Stripe/Twilio; DevOps beyond the basics.
Practicalities
Compensation: USD 2,000-3,300 per month, depending on experience. This is benchmarked to senior local-market rates: roughly INR 1.7-2.8 lakh/month in India.
Setup: contract via Employer of Record · fully remote · 12 months with possible extension, subject to funding.
To apply: send a CV plus links to 1-2 things you've built (repo, demo, or write-up) and 5 lines on how you'd design the chatbot escalation logic → contact@rethinkwellbeing.org.

Work arrangement
Yes