Machine Learning | Computer Vision | Data Science | Reliable end to end AI systems

Thank you for visiting my profile. I solve problems applying cutting-edge AI while never losing touch with the underlying physical meaning. In recent years, my focus shifted towards deep learning based solutions with strong emphasis on using autoML components on the training side and real-time requirements (low latency) on the inference side - especially for computer vision.


You can find some infos regarding my tech stack below:

There exists a plethora of programming languages. My usage looks like this:
- Python: for prototyping & cloud or internal deployment
- C++/Qt: for low-latency & on-premise deployed systems
- Rust: for critical microservices/REST APIs
- html/css/javascript/streamlit: rudimentary web UIs/dashboards
- Matlab,R,SQL,Fortran: used them a lot in my academic past


My machine learning stack converged towards using:
- pandas/scikit-learn/rapids: for "classical" data science/machine learning
- pytorch: for training deep learning models
- mlflow: mainly for logging
- onnx & TensorRT : for deployment of low-latency computer vision pipelines (mostly C++)
- OpenCV: well, the one stop shop for computer vision
- librosa/cuSignal/CuPy for python-based deep signal processing work
- data storage: mainly SQLite, HDF5, PostgreSQL, Redis, various object storage systems (e.g. S3 and others)
- deployment: cloud service provider agnostic (AWS, Azure, GCP, Hetzner, OpenStack, etc.) using docker or proper on-device deployment to squeeze out the last bit of performance