Portrait of Nils Kuhn

Nils Kuhn

I am a Research Scientist at Meta FAIR. I received my M.S. in Electrical Engineering from Stanford University, where I worked with Jeannette Bohg in the Stanford AI Lab on force-aware VLA policies (CoRL 2026). Before that, I interned at the European Space Agency and built continual learning methods for anomaly detection in satellite telemetry (first author, Acta Astronautica 2026). I hold a B.Sc. in Mechatronics and a B.Sc. in Engineering and Business Administration from Leibniz University Hannover. During my undergrad, I completed my thesis under the supervision of Marc Wurz on thin-film insulation, and I worked with Christian Ospelkaus on trapped-ion quantum computing.

My studies were supported by a Stanford Tuition Fellowship and scholarships from the German Academic Exchange Service (DAAD) and the Friedrich-Ebert-Stiftung. I received the Jürgen Ulderup Award and the Wilhelm Launhardt Award for my two bachelor’s degrees.

Feel free to contact me:

Research interests

My research focuses on machine learning for decision-making from noisy data. I am interested in reinforcement learning, world models and predictive modeling. In particular, I study methods that generalize and stay reliable when the data distribution shifts. I have applied these ideas to robot control, satellite operations and trading.

Publications

  • Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them

    C. Parés-Morlans, N. Kuhn, I. Liu, A. Longhini, J. Bohg

    Conference on Robot Learning (CoRL), 2026

    We trace VLA failures in contact-rich manipulation to two root causes, precision failures from flow-matching training and force failures from the structure of force signals, and fix each with a targeted mechanism in FACT, which reaches a 66% average success rate across five real-world tasks versus 41% for the best prior baseline.

  • Validation-Gated Continual Learning for Anomaly Detection in Satellite Telemetry

    N. Kuhn, B. Sánchez Gómez, N. Moreno Blasco, P. Caserman, F. Antonello

    Acta Astronautica, 2026

    We adapt satellite anomaly detectors over time with continual and active learning, deploying each update only after it passes validation, and reach a 92.9% F0.5 score on ESA's telemetry benchmark using under 5% of the original training data.

Projects