Machine Learning Research Scientist · Munich
I develop efficient AI that connects principled learning with real-world deployment.
From Bayesian sparsification and low-bit quantization to physics-guided vision, I work on models that do more within the limits of compute, memory, and energy.
Research & industry experience
Selected research
Physics-driven learning for real-time image dehazing.
Physics shapes training to focus on the most degraded regions. A compact multimodal network restores visibility without adding inference overhead.
Train once for more reliable low-bit deployment across heterogeneous edge accelerators.
Use Bayesian model selection to learn networks that can be pruned more effectively.
Research agenda
Bayesian inference, sparsification, and quantization that connect statistical reasoning to smaller models.
Physics-guided learning and multimodal vision for perception under real-world constraints.
Vision and language models optimized for accelerators, with latency, memory, and energy in view.
Ongoing research extends this agenda to efficient learning and intelligent systems. Further details will follow when the work is ready to share.
The person behind the papers
I bring experience across academic research, automotive systems, optics, and semiconductors. My background in robotics and electrical engineering keeps the full system in view, from learning objectives to the hardware running the model.
More about me ↗Research & engineering notes ↗Let’s connect
Get in touch about efficient AI, research collaboration, or edge deployment.