For a stealth project, we are seeking a Robotics Researcher with a strong foundation in data-driven generalist models to advance our efforts in large-scale, multimodal data collection, curation, and model training for robotics. You will contribute to the development of systems that enable robots to generalize across tasks, environments, and sensor modalities.
Tasks
Responsibilities
- Design and implement pipelines for large-scale robotics data collection and annotation (video, egocentric, 3D, simulation).
- Develop and evaluate generalist models for perception, control, and world modeling using collected data.
- Automate dataset ingestion, cleaning, and integration from heterogeneous sources (simulation, real-world sensors, public datasets).
- Collaborate on benchmarking, dataset standardization, and metadata schema design for robotics datasets.
- Integrate multimodal data (RGB, depth, IMU, tactile) into unified model training workflows.
- Prototype and train models across manipulation, mobility, and navigation domains using state-of-the-art frameworks.
- Contribute to open-source toolkits, evaluation benchmarks, and internal research reports.
Requirements
Requirements
- Strong background in computer vision, machine learning, or embodied AI.
- Experience with model training (transformers, diffusion, or imitation/RL models).
- Proficiency in Python and familiarity with PyTorch, TensorFlow, or JAX.
- Experience working with large datasets (e.g., manipulation, simulation, egocentric).
- Understanding of ROS, Gazebo, Isaac Sim, or similar robotics ecosystems.
- Strong collaborative and documentation skills.