Senior Machine Learning Engineer (Tech Lead), Robot Learning, Loco- Manipulation

Path Robotics is building the future of embodied intelligence with AI-driven systems for robots. They are seeking a Senior Machine Learning Engineer (Tech Lead) to set the technical direction for a new Robot Learning team focused on loco-manipulation in heavy manufacturing, mentor team members, and drive architectural decisions.

Responsibilities:

• Set the ML technical direction for the team, architectural choices on perception, reasoning, and action generation; training methodology; data strategy; the path from research bet to deployed capability
• Own the architectural workstreams that define the team's research and engineering bets — multiple core build streams across action-policy learning, world-model-based supervision, and policy-orchestration interfaces
• Design hybrid physics-ML architectures for the integrated loco-manipulation stack. Manipulation, locomotion, and the whole-body control coupling between them are not separable on legged platforms; sub-millimetre continuous-trajectory precision at the tool requires the whole-body controller to compensate for base motion in real time. Today on fixed bases; tomorrow on mobile platforms. The integration is the hard problem; you own its design
• Own the cross-functional partnerships with hardware teams, domain experts, customer-facing assurance standards, and upstream / downstream teams. Drive a phased deployment strategy that builds production trust over time
• Mentor and shape the team, guide junior and intermediate ICs across software, ML, robotics, and perception backgrounds; establish code-quality standards, review practices, and engineering norms; help identify and attract next hires. You write code throughout — this is a tech-lead role, not a step away from the work

Requirements:

• Ph.D. or Master's degree in Robotics, Mechanical Engineering, Electrical Engineering, Computer Science, or a related field or equivalent experience
• 5+ years of hands-on robot learning experience. You have shipped sim-to-real policies on real robots, across different tasks or platforms
• Demonstrated technical leadership and mentorship. You have made architectural decisions on robot learning systems that others built on, and you have meaningfully shaped the development of more-junior engineers as a tech lead. This can be in academia (leading a lab subgroup, advising students) or industry
• Deep sim-to-real expertise — domain randomisation, system identification, teacher-student distillation, sim-to-online fine-tuning. You can design a transfer strategy for a novel problem
• Full-stack robot learning — fluent across simulation construction, policy training, data collection, real-world deployment, and failure diagnosis
• Physics-informed ML or hybrid control experience — PINNs, neural ODEs, MPC with learned dynamics, process-model-conditioned generation, or similar
• A defensible view on visual-reasoning-centric substrates for grounded spatial / physical reasoning
• Push-back willingness — you can defend a non-obvious architectural commitment under pressure, and change your mind on evidence
• Strong programming skills in Python and C++; production-quality code with reproducibility, testing, and maintainability discipline
• Strong communication skills, able to convey complex technical concepts to a diverse audience
• Demonstrated independent technical authority — you have set technical direction in a tech-lead capacity, leading a research subgroup, owning an architecture across multiple ICs' work, or making the architectural call on a high-profile project
• Edge inference depth (TensorRT, ONNX, edge-class deployment)
• Loco-manipulation experience — locomotion, whole-body control, and manipulation on legged platforms (quadrupeds, humanoids). The coupling between the three is the hard problem; direct hands-on experience or a defensible architectural view both count
• Precision manipulation or surgical robotics — sub-mm accuracy tasks
• Flow-matching action-head design at depth — direct experience with the architectural pattern
• Visual self-supervised representation learning experience on robot or 3D-vision tasks
• Multi-skill workflow or hierarchical policy design — skill sequencing, failure detection, control mode transitions
• Experience building ML capability from early stage — first or second ML engineer on a team, or built a research group's infrastructure from scratch