Senior Software Engineer (ML Infrastructure)

About this role

• You will build the ML platform — the APIs, workers, and control planes that let researchers and robot teams move data and models through the system in a self-serve manner, with the testing and observability that being a dependency implies
• Data Curation & Annotation: Turn raw robot and simulation data into training-ready datasets — selection and filtering of manipulation episodes with synchronized sensor streams; annotation workflows that combine automatic labeling with human-in-the-loop review at throughput; and dataset versioning and lineage strong enough that any model traces back to the exact data that produced it
• Data Pipelines at Scale: Make multi-terabyte dataset operations routine — transformation and assembly, coverage and quality statistics that tell us a training set is good before we spend a cluster-week on it, and read paths that keep GPUs fed
• Simulation & Evaluation: Build the rollout harnesses that evaluate policies in simulation on our GPU cluster; the benchmarks and metrics captured consistently across simulation, real-robot, and teleoperation sources; and the qualification gates a model must pass before it reaches Apollo — automatic, not manual review
• Model Promotion: Build the model store — versioning, metadata, attached evaluation results, lineage — and the promotion path from trained to qualified to deployed on robot, including packaging (ONNX, TensorRT) in partnership with Autonomy
• Developer Experience: Provide the tooling researchers use daily — experiment tracking, training job submission, sweeps, and reproducible container environments. Reduce time from idea to running training job; win adoption by being the fastest path, not by mandate
• Alongside the technical work, you will partner with Autonomy, Data Platform, and TeleOp on dataset and model lifecycle contracts, contribute to the technical direction of these layers, and mentor the engineers around you through code and design review- No single person will have depth in everythi...