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.
Training AI models on videos of drones scrapped from public sources. Everyone in Europe is building ani drone systems and nobody has real data. Lots of companies operating on synthetics only.
Lots of work to automatically filter and process scraped footage into something that will train well.
Senior ML/AI Engineer & Technical Leader with 12+ years building AI/ML systems, scalable backend architectures, and document intelligence platforms. PhD in experimental Physics.
Deep expertise in diffusion models, LLMs, retrieval systems, and real-time pipelines.
Frequent conference speaker (EuroPython 2025, PyCon Lithuania 2024/2025, Data Science Summit). Organizer of AI code and coffee meetup in Warsaw.
Experienced at bridging research → production, mentoring teams, and scaling MVPs to products with significant revenue impact.
We are now moving to a post human economy. When AGI automates all human labour, the consumer i.e. the bulk of humanity stops mattering (economically speaking). It then just becomes Mega corps run by machines making stuff for each other. Resources are then strictly priorities for the machines over everything else. We are seeing this movement already with silicon wafers and electricity.
Sales - especially B2B.
I've got a strong technical background (PhD in Physics) and have been writing code for almost 15 years (most domains, with ML more recently).
I'm also comfortable with public speaking (talking a conferences, pitching etc).
I feel sales is the last piece of the puzzle I'm missing
Senior ML/AI Engineer & Technical Leader with 12+ years building AI/ML systems, scalable backend architectures, and document intelligence platforms. PhD in experimental Physics.
Deep expertise in diffusion models, LLMs, retrieval systems, and real-time pipelines.
Frequent conference speaker (EuroPython 2025, PyCon Lithuania 2024/2025, Data Science Summit). Organizer of AI code and coffee meetup in Warsaw.
Experienced at bridging research → production, mentoring teams, and scaling MVPs to products with significant revenue impact.
Collecting public datasets for training visual AI models to track and target drones.
Drones are real bastards - there's a lot of startups working on anti drone systems and interceptors, but most of them are using synthetic data. The data I'm collecting is designed to augment the synthetic data, so anti drone systems are closer to field testing
Senior ML/AI Engineer & Technical Leader with 12+ years building AI/ML systems, scalable backend architectures, and document intelligence platforms. PhD in experimental Physics.
Deep expertise in diffusion models, LLMs, retrieval systems, and real-time pipelines.
Frequent conference speaker (EuroPython 2025, PyCon Lithuania 2024/2025, Data Science Summit). Organizer of AI code and coffee meetup in Warsaw.
Experienced at bridging research → production, mentoring teams, and scaling MVPs to products with significant revenue impact.
Anonymization of PII data in documents using diffusion models - I'm in the process of reproducing academic papers. The idea is you can replace sensitive information from financial/medical documents with synthetic analogues without visually altering them, so they can be kept/used for AI training
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
GitHub: 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.