Machine Learning Engineer - Robotics (Hugging Face & Isaac Sim)

Overview

We are seeking a Machine Learning Engineer to specialize in training, fine-tuning, and deploying advanced machine learning models within our robotics platform. This position bridges cutting-edge foundational ML models with tangible robotic systems — utilizing the Hugging Face ecosystem to adapt models and datasets tailored for robotics applications, validating them through NVIDIA Isaac Sim, and then deploying onto physical devices.

Key Responsibilities

• Customize and fine-tune foundational models including vision-language models and vision-language-action policies by leveraging Hugging Face tools such as Transformers, PEFT/LoRA, Accelerate, and Datasets.
• Develop comprehensive training workflows for imitation learning and robot policy learning, involving meticulous data gathering, organization, and version control inspired by LeRobot datasets.
• Create and utilize synthetic datasets derived from Isaac Sim, employing domain randomization techniques to facilitate sim-to-real transfer.
• Assess and validate policies and perception models within Isaac Sim and Isaac Lab environments by designing evaluation metrics, implementing test frameworks, and conducting closed-loop simulations.
• Optimize machine learning models for embedded edge GPU deployment, including quantization, model distillation, and converting models to TensorRT or ONNX formats.
• Manage and analyze experimental results using tools like Weights & Biases or MLflow to drive iterative improvements.
• Collaborate with simulation and robotics engineering teams to integrate data generation, model training, simulation evaluations, and physical hardware implementation seamlessly.
• Continuously track developments in physical AI areas such as vision-language-action models, world models, and robot foundation models to rapidly prototype innovative methods.

Qualifications

• Minimum of three years experience in machine learning engineering, with practical involvement in robotics, autonomous systems, or embodied AI.
• Comprehensive, hands-on knowledge of the Hugging Face ecosystem (Transformers, Datasets, PEFT, Accelerate) and proficiency working with the Hub for managing models, datasets, and spaces.
• Advanced skills in PyTorch, with a thorough grasp of transformer architectures, model fine-tuning strategies, and training process dynamics.
• Experience working with NVIDIA Isaac Sim (and preferably Isaac Lab), including environment setup, policy execution in simulation, and utilization of simulation outputs for training and evaluation.
• Strong expertise in Python programming emphasizing clean code practices, unit testing, and reproducibility.
• Comfortable with Linux environments, GPU computing, and distributed or multi-GPU training workflows.

Preferred Experience

• Familiarity with vision-language-action models such as GR00T, OpenVLA, or π0, or working knowledge of robot learning frameworks like LeRobot.
• Experience conducting reinforcement learning within GPU-accelerated simulators such as Isaac Lab or MuJoCo.
• Exposure to sim-to-real transfer techniques including domain randomization and system identification, as well as deployment on actual robotic platforms.
• Knowledge of ROS 2 and executing real-time inference on embedded hardware like NVIDIA Jetson devices.
• Experience with synthetic data creation pipelines, for example Isaac Replicator or Cosmos.
• Contributions to open-source projects in machine learning or robotics communities.