Deep Learning Research Engineer with a path from precision manufacturing through software engineering to ML systems research. I bring hardware intuition, systems thinking, and research rigor to everything I build.
Professional Experience
Machine Learning Research Engineer @ Fraunhofer IIS (Jul 2023 – Present)
Building production ML systems for industrial computer vision on GPU clusters.
- Custom CUDA kernels for image processing, tiled inference, and IoU computation
- Data-parallel training infrastructure: 32 GPUs / 8 nodes, NCCL, mixed-precision, Apptainer
- 5× runtime improvement on C++/OpenMP image analysis via parallelization redesign
- AutoML defect detection for semiconductor manufacturing (production-deployed)
- SAM-based segmentation for industrial waste detection
- NVIDIA DALI integration eliminating I/O bottlenecks at scale
- Zero-shot land cover classification via CLIP on remote sensing imagery
Doctoral Research Scientist @ Technical University of Munich (Oct. 2020 – Jun. 2023)
Compression-aware optimization of deep learning pipelines for edge and satellite systems.
- 19.9% training time reduction via parametric compression framework (3 CNNs, 2 datasets, <1% accuracy loss)
- 25.2% inference preprocessing speedup on FPGA through input compression
- INT8 quantization to FPGA via VitisAI — benchmarked 11 architectures
- Wildfire detection for satellite onboard processing (TensorFlow)
- Compression-based trajectory similarity (Python/C++): 8.9% accuracy improvement over baselines
Research Scientist @ Bundeswehr University Munich (Dec. 2019 - Sep. 2020)
- Published Python/C++ package (pybind11, Boost, >1,000 LOC C++ core) for raster data processing and entropy-based similarity
- Linux build pipelines for C++ based Python packages with Make, Boost, pybind11
Student Research Assistant @ German Aerospace Center (Feb 2019 – Jul. 2019)
- Change detection framework for satellite imagery: 70,000 km² processed
- 36% runtime reduction via memoization and overhead elimination
- Satellite data processing pipelines
Software Engineer (intern) @ Fraunhofer IIS (Feb 2017 – Jul. 2017)
- Built performance monitoring modules for a distributed cinema rendering system, identifying hardware and software bottlenecks across nodes.
Earlier Career
Optiplan · Technical Staff · 2012 – 2014
Arvai Plastics · Toolmaker (Apprenticeship + Journeyman) · 2007 – 2012
Started in precision manufacturing — CNC machining, injection mold tooling. This background gives me an intuitive understanding of hardware constraints, tolerances, and production systems that directly informs my approach to deploying ML at the edge.
Education
Doctor of Natural Sciences in Aerospace and Geodesy @ Technical University of Munich (Oct. 2020 - May 2024)
- Thesis: Aspects of Algorithmic Information Theory in Spatial Machine Learning
- Focus: Optimization of data-driven pipelines using compression to increase performance and scalability.
Diplom-Ingenieur @ Salzburg University of Applied Sciences (Sep. 2017 - Oct 2019)
- Program: Information Technology & Systems Management
- Thesis: Supervised and Unsupervised Data Mining Methods in Remote Sensing
Bachelor of Science @ Salzburg University of Applied Sciences (Sep. 2014 - Jul 2017)
- Program: Information Technology & Systems Management
- Thesis: Performance Data Collection in a Distributed System for Rendering Cinema Movies
Technical Stack
| Domain | Technologies |
|---|---|
| GPU Computing | CUDA, C++20, pybind11, OpenMP, Boost |
| Distributed Training | PyTorch DDP, NCCL, AMP, Slurm (up to 32 GPUs / 8 nodes) |
| Inference & Deployment | TensorRT, VitisAI, INT8/FP16 quantization |
| Data Pipelines | NVIDIA DALI, custom preprocessing kernels |
| Models | CLIP, SAM, defect detection, remote sensing classification |
| Infrastructure | Docker, Apptainer, Linux, HPC cluster management |
Certifications
| Course | Provider | Year |
|---|---|---|
| Introduction to Parallel Programming with MPI | NHR@FAU | 2025 |
| Node-Level Performance Engineering (NUMA, SIMD) | LRZ | 2025 |
| Fundamentals of Accelerated Computing with CUDA C/C++ | NVIDIA | 2025 |
| Fundamentals of Accelerated Computing with CUDA Python | NVIDIA | 2025 |
| GPU Programming Workshop (OpenACC, Nsight Profiling) | LRZ | 2025 |
| Parallel Programming of HPC Systems (OpenMP, MPI) | LRZ | 2023 |
Publications
18 peer-reviewed publications including ACM Computing Frontiers, IEEE MLSys, IGARSS, IEEE JSTARS.
→ Full publication list · Google Scholar
Teaching & Mentoring
Fraunhofer IIS — Internal Workshops (2023 - Now)
- HPC cluster infrastructure & job scheduling
- Multi-node distributed training (DDP/NCCL)
- Kubernetes/Kubeflow for ML workloads
Technical University of Munich (2020 – 2023)
Teaching assistant for C++ programming and spatial data science —
5 semesters, covering OOP, templates, STL, memory management,
multi-threading, MPI. (~5 hrs/week)
Thesis Supervision
- Cross-Modal Pseudo-Labeling and Label Expansion for Domain-Adaptive Semantic Segmentation (M.Sc., 2026)
- Forest Fire Detection from Satellite Imagery (M.Sc., 2023)
- Analysis of the Potential of Quantum Machine Learning for Remote Sensing (M.Sc., 2022)