Top Tech Jobs & Startup Jobs

Design, train, and validate multimodal, action-conditioned world models (video, state, action, language) for humanoid robots. Develop learned-simulator evaluation, generate synthetic rollouts for sim-to-real and policy training, build data pipelines from fleet telemetry, run large distributed training and scaling studies, and collaborate with pretraining, RL, and manipulation teams to integrate models into policy and safety workflows.

Lead development and integration of production-grade, deterministic control software for humanoid robots. Design real-time control architectures, implement high-performance control loops, industrialize algorithms, ensure safety and reliability, run automated testing (simulation and hardware-in-the-loop), manage releases, profile and optimize performance, and mentor engineers.

Develop and maintain embedded Linux platform software and core subsystems for humanoid robots, integrating sensors and safety systems, building kernel/device drivers, ensuring performance and reliability, leading cross-functional projects, and contributing infrastructure and safety practices for real-world deployments.

Build and maintain physics-based simulation environments and RL infrastructure for dexterous manipulation across industrial, service, and home domains. Design Gym-style observation/action/reward spaces, mitigate reward hacking, reduce sim-to-real gap, validate physics against real data, integrate high-fidelity 3D assets, optimize simulation throughput, and support large-scale GPU-accelerated training.

Build and operate a petabyte-scale data platform and Capability Factory: ingest, curate, preprocess, and version robot telemetry; design scalable batch/stream pipelines, orchestration, compute and storage; collaborate with ML teams on MLOps and evaluation workflows; build data warehouse and BI dashboards; ensure data quality, access control, and production ownership including on-call incident triage.

Design, simulate, and validate bipedal and wheeled robot platforms using physics simulation and reinforcement learning. Own URDF models and the CAD->URDF->simulation pipeline, develop benchmarks for locomotion/manipulation, analyze dynamics and actuator sizing, and provide quantitative feedback to hardware, control, and perception teams.

Lead end-to-end industrial product delivery across hardware, software, AI, and system integration. Own prioritization and roadmaps for engineers dedicated to industrial products, partner with functional leads and Ops for manufacturing readiness, and ensure pilots transition to scalable, reliable deployments meeting SLAs.

Own and improve the build system and CI infrastructure for a multi-target monorepo spanning firmware, robot software, simulation, and cloud. Build and maintain GitHub Actions pipelines, instrumentation and dashboards for CI telemetry, enforce merge/PR policies, manage remote build cache and dependency mirrors (apt, PyPI), and improve developer local-build experience to shorten feedback loops and keep main healthy.

Develop and train language-vision conditioned manipulation policies with reinforcement learning in simulation and real robots. Build diverse simulated task suites, collect teleoperation trajectories for behavior cloning, establish real-world RL training pipelines, and research sim-to-real transfer methods.

Build and maintain device software and data pipelines for handheld gripper data collection devices. Implement sensor integration and temporal synchronization, recording/packaging episodes, validation tooling to detect data faults, operator-facing controls, and on-site support. Debug hardware integration issues and collaborate with hardware, AI, and data teams to ensure high-quality training datasets.