Staff MLOps Engineer

Apptronik•Austin, TX

3d•Onsite

About The Position

Apptronik is a human-centered robotics company developing AI-powered robots to support humanity in every facet of life. Our flagship humanoid robot, Apollo, is built to collaborate thoughtfully with people, starting with critical industries such as manufacturing and logistics, with future applications in healthcare, the home, and beyond. We operate at the cutting edge of embodied AI, applying our expertise across the full robotics stack to solve some of society's most important problems. You will join a team dedicated to bringing Apollo to market at scale, tackling the complex challenges like safety, commercialization, and mass production to change the world for the better. Apptronik is seeking a Staff MLOps Engineer to own the technical direction of our MLOps platform — the system of record for datasets, experiments, model artifacts, and serving paths that connects teleoperation data collection on one side to deployed autonomy on Apollo on the other. In this role, you will set the architecture for the platform layer above the training cluster: dataset lifecycle, experiment tracking, model registry, evaluation harnesses, and the serving / packaging path that delivers trained policies to robots in the field. You will lead by influence across MLOps, Autonomy, Data Platform, and TeleOp — establishing the standards, contracts, and tooling that turn one-off research code into a repeatable, auditable pipeline from data to deployed model. This is a hands-on technical leadership role, not a management position; you will be a primary contributor while mentoring the engineers around you, and partnering closely with the Training Infrastructure engineer who owns the cluster layer beneath the platform.

Requirements

Deep proficiency in Python and at least one systems-level language (Go, Rust, or C++), with demonstrated ability to make and defend architectural tradeoffs in production ML platforms
Proven experience owning and delivering an MLOps platform end-to-end — dataset lifecycle, experiment tracking, model registry, evaluation, and serving — at a company that ships models to production
Expertise across the model lifecycle: dataset versioning (DVC, LakeFS, Delta, or equivalent), experiment tracking (MLflow, W&B, Determined), model registry, and policy serving
Strong background designing service-oriented systems on Kubernetes; comfortable with the contract between platform APIs and underlying compute infrastructure
Experience defining evaluation and qualification frameworks for ML models where the cost of a regression is high (robotics, safety-critical, or production-customer-facing)
Experience leading technical projects end-to-end: architecture, implementation, validation, and iteration
Demonstrated ability to lead by influence across teams — setting standards that other engineers adopt voluntarily, and mentoring engineers around you
Proficiency with cloud infrastructure (AWS, GCP, or Azure), Docker, Git, and modern CI/CD workflows

Nice To Haves

Experience deploying ML models to edge or embedded targets (on-device inference, ONNX Runtime, TensorRT, robot fleets)
Experience with RL training and evaluation infrastructure for embodied agents (rollout workers, replay buffers, sim-eval harnesses)
Familiarity with humanoid robotics, dexterous manipulation, or teleoperation data domains
Experience with simulation-in-the-loop evaluation (IsaacSim, MuJoCo, or equivalent)
Familiarity with policy gating, shadow deployments, or staged rollout strategies for autonomy
Open-source contributions to MLOps platform tooling (MLflow, BentoML, KServe, Ray Serve, etc.)

Responsibilities

Own the technical direction for the MLOps platform — define subsystem interfaces, drive architecture decisions, and establish engineering standards for how datasets, experiments, and models move through Apptronik's systems.
Serve as the primary technical point of contact for Autonomy, Data Platform, and TeleOp on all matters of model lifecycle and platform contracts.
Design and operate the dataset layer end-to-end — versioning, lineage, splits, and labeling-integration handoff.
Ensure every trained model can be traced back to the exact data and code that produced it.
Build and operate a first-class model registry — versioned artifacts, metadata, evaluation results, lineage, and approval workflows.
Define the promotion path from "trained" to "qualified" to "deployed to robot."
Define the offline benchmarks, simulation rollouts, and policy-gating harnesses that any model must pass before reaching Apollo.
Develop the metrics framework that the autonomy team trusts to gate releases.
Own the path from registered model to running inference on Apollo — packaging (ONNX, TensorRT, torch.compile), versioning on-robot, rollback, and observability of deployed policy behavior.
Coordinate with Connect and Data Platform on the deploy-and-telemetry seam back from the fleet.
Mentor mid-level and senior engineers on the MLOps team through code review, design review, and direct collaboration.
Partner with the Training Infrastructure engineer on the cluster/platform contract, and influence research workflows across Autonomy to standardize on the platform's primitives.