At Aptiv, I deploy and performance-test neural-network inference for driver and cabin monitoring, using ONNX pipelines and TIDL on TI TDA4. I own release integration and lead two senior engineers, with improvements to release workflows reducing estimated release effort by approximately 30%.
I developed C++ camera-frame capture and shared-memory transport at Aptiv, and integrated an IPC framework that reduced communication latency by approximately 50%. I also worked on body-keypoint features, camera calibration, and blockage detection across desktop and embedded platforms.
My PhD research at AGH University of Krakow combined neural-network training with embedded implementation for adaptive sensor-data compression. I replaced approximately three seconds of parameter search with one millisecond of inference, implemented the compression approach in embedded C++, and published peer-reviewed research on Smart Grid signal processing and compression.
