• Own major features end-to-end. From design discussions to production deployment. • Build across the stack. Frontend, backend, and everything in between. • Instrument and learn. Use metrics and user feedback to refine what we build. • Shape the foundation. Influence architecture, tools, and engineering practices. • Keep security and ethics front and center. We’re handling sensitive healthcare data — trust is everything. • We’re tackling complex engineering problems — building intelligent systems that can understand messy, real-world healthcare data, automate workflows, and communicate with empathy and reliability.

• Own major features end-to-end. From design discussions to production deployment. • Build across the stack. Frontend, backend, and everything in between. • Instrument and learn. Use metrics and user feedback to refine what we build. • Shape the foundation. Influence architecture, tools, and engineering practices. • Keep security and ethics front and center. We’re handling sensitive healthcare data — trust is everything. • Tackle complex engineering problems — building intelligent systems that can understand messy, real-world healthcare data, automate workflows, and communicate with empathy and reliability.

• Own Autocomplete end-to-end: product feel, model behavior, latency, reliability, quality, and iteration velocity. • Build the benchmark that matters: not a vanity leaderboard — a real eval harness that tracks what users actually experience (acceptance rate, edit distance, time-to-accept, latency, regressions by language/project type). • Run tight experiment loops: tweak prompts, decoding, routing, caching, context packing, and ranking — then measure, learn, repeat. • Form strong opinions: FIM vs next-edit, speculative decoding, reranking, context window strategy, model routing, and how to trade off quality vs speed without lying to yourself. • Work directly with model creators: test new models early, give actionable feedback, and ship upgrades fast when something is truly better. • Ship constantly: small improvements weekly, bigger improvements monthly, and a compounding moat over time.

• Ship an internal release of a new feature in week 1 • Get plugged into the user community and start chatting with users in month 1 • Own a product area and measure feature performance in month 1 • Be flying across the code base with comfort, confidence, and purpose by month 3