Location: Warsaw, Poland (CET) Remote: Yes (EU/US overlap fine) Willing to relocate: Yes Work authorization: UK/EU Technologies: Python, PyTorch, C++, CUDA/

Location: Warsaw, Poland (CET) Remote: Yes (EU/US overlap fine) Willing to relocate: Yes Work authorization: UK/EU Technologies: Python, PyTorch, C++, CUDA/TensorRT, NVIDIA Jetson, object detection & tracking, detection theory, ML evaluation, CV data pipelines Résumé/CV: https://piotrgryko.com/assets/pdf/Piotr-Gryko-CV.pdf https://piotrgryko.com/assets/pdf/Piotr-Gryko-CV.pdf GitHub: https://github.com/pgryko https://github.com/pgryko Email: piotr.gryko@gmail.com Physicist turned ML engineer (PhD, Imperial), 12+ years in engineering, 3+ in production computer vision. I build measurement-first CV systems: evaluation, GPU/edge inference, and training-data pipelines. - Counter-UAS, defense-tech client: building real-time detection of few-pixel aerial targets in compressed video on Jetson under a hard latency budget, using stabilization, CFAR, and track-before-detect rather than deep models. Wrote the Pd/FROC/FAR evaluator with confidence intervals and reproducible run configs before the detector. - Document-AI company (3 years): built GPU inference services plus the labelling and evaluation platform (versioned golden datasets, per-class F1, calibration, mislabel detection). Used it to move a new classifier from shadow mode to primary with divergence logging. - Own production data engine: 220k publicly posted videos, 118k VLM analyses, embedding-based keyframe selection, CVAT-assisted annotation, YOLO/COCO/VOC export. Its evaluation caught a detector's accuracy collapsing from 0.875 to 0.375 and a migration costing 2.6x the "cost-neutral" estimate. Seeking: senior/staff IC in computer vision/perception, edge/GPU inference, or ML evaluation/data infrastructure. Full-time or B2B contract; open to defense, sensing, robotics, and scientific ML.