Make the agents that run healthcare's front desk smarter, faster, and more reliable. Own the models, evaluation, and training loops behind millions of real conversations, and turn what happens on the phone every day into a system that keeps getting better. WHAT YOU'LL DO You'll own the machine learning behind Dodo's agents, from the models in the call path to the evaluation that tells us whether they work. - Improve how our agents understand callers, choose actions, and complete bookings across voice, text, and web - Build evaluation pipelines and judges that measure agent quality on real conversations, catch regressions, and gate every model and prompt change - Fine-tune, distill, and route models to hit the latency and cost budgets of real-time voice while improving quality - Build the data loops that turn conversation outcomes, practice feedback, and corrections into training signal - Own speech and voice quality where it meets the model: transcription, turn-taking, interruption handling, and latency - Work with product engineers to ship model improvements into production and measure their effect on bookings and customer outcomes - Set the direction for how Dodo uses models as the space changes month to month WHO YOU ARE You've shipped ML into production systems people depend on, and you care about measured results more than demos. - 3+ years building and deploying machine learning or LLM systems in production - Deep experience with LLM evaluation, fine-tuning, prompt and model routing, or speech systems - Strong engineering fundamentals: you write production code and own what you ship - Rigorous about measurement, and comfortable designing evaluations when none exist - Pragmatic about latency, cost, and reliability in real-time systems - Comfortable working with messy real-world conversation data and the privacy obligations of healthcare - Excited to work in person with a small team of exceptional engineers in San Francisco