Gerald Mwangi

My name is Gerald Mwangi. I hold a Physics degree specializing in Theoretical Physics from Heidelberg University, where I developed strong analytical and problem-solving skills through theoretical and experimental work.

At Eye Vision Technologies, I applied my theoretical background to industrial innovation in mathematical image processing. I led the redesign and optimization of the EyeSpector software's filter area, significantly improving performance and stability.

I pursued a PhD at the Heidelberg Collaboratory for Image Processing (HCI) under Prof. Christoph S. Garbe. During this time, I implemented advanced algorithms for multimodal optical flow and Structure-from-Motion (SfM) pipelines for industry partners. I developed a novel optimization algorithm based on an extension of the Principle of Least Action (PLA), achieving substantial performance improvements over conventional methods.

As Lead Software Developer at ASE/Vitronic (ase-gmbh.com, vitronic.de), I established the software department and developed Numberfinder—the C++ core of NumberCheck, an automated video analysis system for cargo trains and trucks. This system performs object recognition and OCR to identify wagons/containers and extract UIC/BIC codes.

I led a major architectural transformation from monolithic to distributed microservice architecture in Rust with C++/Rust plugins. I integrated custom OCR neural networks and deployed YOLO for vehicle code recognition, achieving a 10x speed increase and improved robustness. I also managed talent acquisition, established agile processes, and implemented CI/CD pipelines with GitLab/Jenkins and Nextcloud.

Beyond software development, I'm an avid guitarist and creator of GuitarMidi-lv2 (github.com/geraldmwangi/GuitarMidi-LV2), a real-time polyphonic guitar-to-MIDI plugin using a neural net for synthesizer control using any electric guitar.