This is a hands-on engineering role for someone who has built and deployed learning systems that interact with the real world - not just trained models in isolation.
What you’ll be working on
- Building and deploying Vision-Language-Action (VLA) and robotics foundation models on real robotic platforms
- Developing end-to-end learning pipelines from data collection → training → evaluation → real-world deployment
- Working directly with robotic hardware (manipulation systems / mobile robots depending on project scope)
- Designing large-scale data pipelines from multimodal robot sensor streams (vision, depth, proprioception, action logs)
- Running structured experimentation across architectures, datasets, and training strategies for physical AI systems
- Improving Sim2Real transfer and closing the gap between model performance in simulation and real-world environments
- Optimising inference and latency for real-time robotic control loops
- 3 years experience in ML/robotics/computer vision with at least some exposure to real-world robotic systems
- Hands-on experience with VLA, robotics policy learning, RL, imitation learning, or multimodal transformer models
- Experience working across the robotics stack (e.g. ROS, simulation tools such as Isaac / Gazebo / MuJoCo, sensor integration)
- Strong track record of data-driven experimentation (not purely intuition-led iteration)
- Experience building or owning data pipelines for ML systems (collection, filtering, labelling, evaluation)
- Comfortable working close to hardware and debugging real-world system behaviour
- Strong Python and deep learning framework experience (PyTorch preferred)
- Experience deploying models directly onto robots (arms, mobile robots, drones, or similar)
- Exposure to large-scale training (multi-GPU / distributed training)
- Experience with Sim2Real transfer or domain randomisation
- Background in transformer-based robotics policies or foundation models
- Experience with real-time systems or embedded deployment constraints
- You’ll work directly on embodied AI systems operating in the real world, not just simulation research
- High autonomy across model design, data strategy, and deployment
- Opportunity to shape core robotics intelligence systems from the ground up
- Strong focus on shipping systems that work outside the lab
