👨💻 Machine learning engineer and researcher.
🤖 Always curious about how things move, learn, and connect.
M.Sc. Informatics (TUM)
Researching at the intersection of machine learning, vision, and scientific computing,
with focus on 3D perception and physics-informed learning across domains like autonomous driving, space applications, and computer vision.
-
Machine Learning Team Lead - WARR DEDRA Science Group (TUM)
Leading the Machine Learning Team of TUM’s largest student space organization.
Developed physics-informed models for space debris impact prediction using ion-based sensor data.
Contributed to a satellite mission launched in 2024, following earlier work as an Embedded Systems Engineer. -
Research Assistant - TUM Autonomous Motorsport (Chair of Automotive Technology, FTM)
Developing 4D radar-only odometry and 3D LiDAR perception for autonomous racing at speeds > 250 km/h.
Achieved state-of-the-art results beyond current benchmarks; preparing for publication in 2026 and competing at the Autonomous Racing League (A2RL). -
Research Intern - Computer Vision Lab (Prof. Daniel Cremers, TUM)
Conducted research on multimodal model merging using subspace parameter compression,
achieving state-of-the-art performance with reduced compute. Publication planned for 2026. -
Machine Learning Engineer - BMW Group
Designing and deploying multimodal ML models that improve perception and simulation performance.
Built scalable ML-Ops infrastructure on AWS SageMaker, optimizing workflows and cutting operational cost across research teams. -
Research Intern - Machine Learning Lab (Prof. Stephan Günnemann, TUM)
Investigated graph neural networks (GNNs) and uncertainty quantification, leading to new insights on ensemble-like behavior.
Authored a seminar paper reviewing Vision Transformers, focusing on robustness and architectural trade-offs.
→ Uncertainty-Quantification-on-GNN-using-Stochastic-Centering
→ Review-on-Vision-Transformers -
Bachelor’s Thesis - Radar Deep Learning for Space Debris Detection (Actlabs × TUM)
Built a satellite-mounted radar detection system using deep learning on simulated I/Q signals.
Demonstrated feasibility of space-borne debris classification with neural radar processing.
→ Deep-Learning-Based-Space-Debris-Classification
- 3D Reconstruction using Machine Learning (Prof. Dai)
- Diffusion-Based Policy Learning in Robotics (Prof. Bäuml)
Core Domains:
Machine Learning · Deep Learning · Computer Vision · Radar & LiDAR Perception · SLAM · 3D Reconstruction · Model Merging · Physics-Informed AI
Stack & Tools:
Python · C++ · PyTorch · ROS2 · Docker · AWS SageMaker · Hydra · Weights & Biases · Git · Slurm · Matlab
- M.Sc. Informatics, Technical University of Munich
- B.Sc. Informatics, Technical University of Munich
- German Federal Scholarship · German Physical Society Award
Check out my pinned repositories below for selected projects and implementations.