About Kin

Healthcare is still a black box for most Americans. Kin turns any medical visit into momentum and becomes the default patient AI layer that works in any exam room today: patient-controlled, privacy-forward, and designed to make care feel less confusing and more doable in the moments that count.

We are a physician-led team of repeat entrepreneurs with multiple successful exits, supported by world-class capital and strategic partners (including the former co-CEOs of GoodRx). We’re hiring a Head of Data & AI to own the AI systems that make Kin useful, trustworthy, and continuously improving across the patient journey.

Why this role matters

Kin’s promise depends on turning messy, emotional, real-world care into clear next steps patients can trust. Our next frontier is not another chatbot; it is a living, provenance-aware understanding of a person’s care journey. One that can connect visits, symptoms, medications, labs, follow-ups, and social context without asking the patient to become the system integrator.

You’ll own the design and operation of Kin’s AI systems end-to-end: LLM orchestration, data foundations, evals, monitoring, and production workflows. Reporting directly to our CTO, Kyle Alwyn, you’ll turn ambiguous healthcare and operational problems into reliable systems that learn from real care, preserve where every claim came from, and improve every week.

What you will be building at Kin

• You’ll own Kin’s AI system architecture. Design the orchestration layer, retrieval patterns, model interfaces, data flows, guardrails, and human-in-the-loop workflows that power patient-facing and internal AI experiences — and quietly turn fragmented care artifacts into coherent context.
• You’ll build the evals and observability backbone. Define what “good” means, turn that into repeatable evaluation systems, and make quality visible across product, clinical, and operational workflows. You’ll know when the system is getting better, when it is drifting, and where it needs intervention.
• You’ll turn messy care problems into production workflows. Visits, summaries, follow-ups, care context, clinical language, operational edge cases, and patient goals all collide in the real world. You’ll design systems that handle that complexity without making the product feel complex.
• You’ll make data a product advantage. Create the structures, feedback loops, and operating practices that let Kin learn from real usage while respecting patient control, privacy, and trust. The right person will see the shape of a new kind of patient memory here: source-aware, longitudinal, and useful without being noisy.
• You’ll help define how an AI-native team operates. Build the workflows, tools, automations, and internal systems that let a small team move with leverage. You won’t just use AI to build faster; you’ll shape how the company builds, measures, and improves AI systems.

What we're looking for

We're building for people navigating the scariest moments of their lives. That shapes everything about who we hire, how we build, and what we ship.

• You’ve built AI systems that survived contact with reality. You’ve taken LLM-powered products or data systems from prototype to production and learned what breaks: brittle prompts, silent regressions, incomplete context, bad feedback loops, unreliable outputs, unclear ownership, and systems no one can actually monitor.
• You know evals are a product surface. You can design qualitative and quantitative evaluation systems, combine automated checks with expert review, and translate ambiguous product quality into measurable signals that teams trust.
• You think in systems, not prompts. Models matter, but you know the real leverage comes from data shape, context engineering, orchestration, latency, cost, monitoring, fallback behavior, review workflows, and the product experience around the AI.
• You can operate across patient, clinical, and business contexts. You’re comfortable working with clinicians, product, engineering, and operations to understand what the system must do, where it must be careful, and how to make tradeoffs when correctness, usefulness, speed, and trust are all in tension.
• You have strong data judgment. You know how to design schemas, pipelines, instrumentation, and feedback loops that compound over time. You care about data quality because you’ve seen how quickly bad data becomes bad product behavior — and because the most important systems are the ones that can answer “what changed, when, why, and how do we know?”
• You’ve already rewired how you build. You use LLMs and agentic tools in your actual day to day, not as experiments. You have real opinions about where they help and where they get in the way.