michalwysocki.com

Michał Wysocki is a machine learning engineer whose work focuses on building systems that model and predict real-world behavior. His professional interests span applied machine learning for autonomous systems and counter-unmanned aircraft, alongside research into world models and representation learning. Since September 2026, he has served as a Junior ML Engineer at Hertz New Technologies, working with multimodal detection, computer vision, prediction, and sensor data.

His portfolio presents selected engineering and research projects. One ongoing project explores path consistency for latent world models, using a self-supervised mechanism to reduce disagreement between direct, horizon-conditioned predictions and autoregressive rollouts. Another investigates drone audio detection through a Random Forest classifier trained on MFCC features. The project highlights a practical sim-to-real generalization gap: strong held-out results did not carry over to real recordings, where ambient noise was often misclassified.

Wysocki shares project code and professional information through his website, GitHub, and LinkedIn. He welcomes contact about research, engineering, and interesting technical problems. The portfolio communicates a blend of practical system-building and research-oriented machine learning.