SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Apptronik is building Apollo, a general-purpose humanoid robot powered by physical AI. The company is scaling production and deployment across manufacturing, logistics, healthcare, and beyond. This role focuses on building the ML infrastructure platform that transforms raw robot and simulation data into deployed autonomy at multi-terabyte scale.
You will own the design and implementation of self-serve platform services and pipelines that carry data from collection through curation, training, and evaluation to qualified models running on real hardware. The platform serves researchers and engineers across MLOps, Autonomy, Data Platform, and TeleOp teams daily.
Key responsibilities include:
**Data Curation & Annotation**: Build workflows to turn raw robot and simulation data into training-ready datasets. This includes selection and filtering of manipulation episodes with synchronized sensor streams, automatic labeling combined with human-in-the-loop review at scale, and dataset versioning and lineage that traces every model back to its exact source data.
**Data Pipelines at Scale**: Make multi-terabyte dataset operations routine through transformation and assembly, quality statistics that validate training sets before expensive cluster runs, and optimized read paths that keep GPUs fed.
**Simulation & Evaluation**: Build rollout harnesses that evaluate policies on GPU clusters, capture consistent benchmarks and metrics across simulation and real-robot sources, and implement automatic qualification gates for model promotion.
**Model Promotion**: Develop the model store with versioning, metadata, evaluation results, and lineage tracking. Build the promotion path from trained to qualified to deployed, including packaging (ONNX, TensorRT) in partnership with the Autonomy team.
**Developer Experience**: Provide daily tools for researchers—experiment tracking, training job submission, hyperparameter sweeps, and reproducible container environments. Optimize the path from idea to running training job.
You will partner with cross-functional teams on dataset and model lifecycle contracts, contribute to technical direction, and mentor engineers through code and design review. This is a hands-on role on a small, high-impact team building first-party platform services alongside open-source and commercial tooling.
