Friedrich Wilke Grosche
Research Engineer · Control & Simulation · Computer Vision · ML
Physics MSc (ETH Zürich) with 3 years of research engineering at EPFL CVLab, specialising in control theory, computer vision, and machine learning. Proven track record translating complex physical systems into tractable optimisation problems and production-grade implementations.
Experience
Research EngineerSep 2022 – Sep 2025
EPFL Computer Vision Lab (CVLab) — Lausanne, Switzerland
Time-optimal MPC for fixed-wing drones — Trajectory optimisation and MPC for F5B competition (100 m/s). Differentiable aerodynamic surrogate from wind tunnel, CFD and flight data, cutting force evaluation from ~8 h to ~1 ms. Angular-rate correction resolving integration instability.
Multi-view camera localisation — Led camera calibration suite; automated axis-alignment via RANSAC plane fitting and height priors.
GNN-based people tracking — Data augmentation and training of graph neural network for long-term multi-view tracking with occlusions (ByteTrack / YoloX).
SCOUT annotation pipeline — Web-based annotation tool for 25-camera dataset; automated tracklet merging reduced annotation time by ~90%.
Cardiac catheter simulation — Cosserat rod + SDF collision model for real-time virtual surgery.
Multi-view camera localisation — Led camera calibration suite; automated axis-alignment via RANSAC plane fitting and height priors.
GNN-based people tracking — Data augmentation and training of graph neural network for long-term multi-view tracking with occlusions (ByteTrack / YoloX).
SCOUT annotation pipeline — Web-based annotation tool for 25-camera dataset; automated tracklet merging reduced annotation time by ~90%.
Cardiac catheter simulation — Cosserat rod + SDF collision model for real-time virtual surgery.
Education
MSc in Physics
ETH Zürich
2019 – 2022 · Thesis: Model-Agnostic Explainability for ML Regression
BSc in Physics
ETH Zürich
2015 – 2019 · Thesis: Flux Gate Optimisation for Active Magnetic Shielding
A-Levels
Hills Road Sixth Form, Cambridge
2013 – 2015 · Physics, Maths, Further Maths, Chemistry, German
Publications
Engilberge, Wilke Grosche, Fua. No Identity, no problem: Motion through detection for people tracking. arXiv 2411.16466, 2024.
Engilberge et al. Unified People Tracking with Graph Neural Networks. arXiv 2507.08494, 2025.