I work on physical AI at Intel, mostly training robot policies from human demonstrations and making them run on edge hardware. Before that I spent several years on visual anomaly detection, starting with my PhD at Durham University and later building Anomalib.
LinkedIn · Google Scholar · ORCID
- Physical AI Studio: an imitation learning framework for robots. You record demonstrations, train a policy (ACT, SmolVLA, Pi0.5, XR0, MolmoACT2, RLDX-1 and anything in LeRobot) and export it to OpenVINO, ONNX or ExecuTorch.
- physicalai: the runtime for deploying those policies. It handles cameras, robot interfaces, inference and the control loop, and currently supports SO-101, Trossen WidowX-AI, Seeed Studio B601 and many other arms.
I gave a talk on this stack at PyTorch Conference Europe 2026: Full-Stack PyTorch Robotics VLA, from Data to Edge via ExecuTorch/OpenVINO.
- Anomalib: a deep learning library for visual anomaly detection, which I created and maintained at Intel.
- Geti: Intel's platform for training computer vision models with less data.
- GANomaly and Skip-GANomaly: code for my papers. The models are now part of Anomalib.
- FEVER-OOD: Free Energy Vulnerability Elimination for Robust Out-of-Distribution Detection. ICCV 2025.
- Beyond Academic Benchmarks: Critical Analysis and Best Practices for Visual Industrial Anomaly Detection. CVPR Workshops 2025.
- Divide and Conquer: High-Resolution Industrial Anomaly Detection via Memory Efficient Tiled Ensemble. CVPR Workshops 2024.
- Anomalib: A Deep Learning Library for Anomaly Detection. ICIP 2022.
- Towards Automatic Threat Detection: A Survey of Advances of Deep Learning within X-ray Security Imaging. Pattern Recognition 2022.
- Skip-GANomaly: Skip Connected and Adversarially Trained Encoder-Decoder Anomaly Detection. IJCNN 2019.
- GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training. ACCV 2018.
The full list is on Google Scholar.
I'm happy to talk about robot learning, edge deployment or anomaly detection, and I'm open to research collaborations. LinkedIn is the best way to reach me.