Machine Learning Engineer, Robotics
Job Description
Path Robotics’ Robotics team is hiring a Machine Learning Engineer to apply machine learning to robot planning, scanning, and actions in real-world environments. This role focuses on turning robotics problems into well-defined ML tasks, training and validating models with robotic-cell data, and deploying end-to-end capabilities on live robots.
Responsibilities
- Translate open-ended challenges in planning, scanning, and robot behavior into structured machine learning problems.
- Train models that work alongside classical robotics methods, including learned costs and initial guesses for trajectory optimization, predictors for plan quality and feasibility, and learned collision models.
- Develop viewpoint planning policies that decide where and how to scan, including predicting scan quality before the robot proceeds.
- Apply imitation learning, reinforcement learning, and generative model approaches such as diffusion and flow matching when they outperform classical baselines.
- Build production-ready machine learning pipelines, including dataset creation from fleet logs and simulation, model training, and metric tracking.
- Evaluate models in the loop, using offline metrics to estimate robot gains, performing failure-mode analysis, and calibrating confidence to determine when the system should defer to classical methods.
- Own models end to end, from prototype through deployment on live robotic cells, including low-latency inference and ongoing monitoring.
Requirements
- MS or PhD in Computer Science, Robotics, Machine Learning, or a related field.
- 2+ years of applied machine learning for robotics or autonomy in research or industry.
- Fluent in Python with PyTorch or JAX, with experience training, evaluating, and deploying deep learning models.
- Strong foundations in linear algebra, 3D geometry, kinematics, and optimization.
- Experience running methods on real robot hardware.
- Ability to explain model outcomes and limitations clearly, enabling planning, perception, and field teams to know when to rely on model outputs.
Technologies
- Python, PyTorch, JAX
- C++, ROS 2
- CUDA, TensorRT, ONNX
- Isaac Lab, MuJoCo
Preferred Qualifications
- C++ experience and deploying models into a C++ or ROS 2 production stack.
- Experience with motion planning, trajectory optimization, or collision checking.
- Active perception approaches, including next-best-view or informative path planning, or 3D point cloud learning.
- Uncertainty estimation and calibration, including Bayesian and information-theoretic planning.
- GPU inference optimization using CUDA, TensorRT, and ONNX.
- Simulation and sim-to-real transfer with Isaac Lab and MuJoCo.
- Publications at RSS, CoRL, ICRA, IROS, NeurIPS, ICML, or similar venues.
Benefits
- Daily free lunch to stay fueled and connected with the team.
- Flexible PTO.
- Comprehensive medical, dental, and vision coverage.
- 6 weeks fully paid parental leave, plus an additional 6 to 8 weeks for birthing parents (12 to 14 weeks total).
- 401(k) retirement plan through Empower.
- Generous employee referral bonuses.
Location and Experience Level
Location: Columbus, OH (onsite). Minimum experience: 2 years.