I build camera-driven AI that survives the real world. My specialty is productionizing computer-vision for CCTV/NVR environments—airports, retail, schools, and manufacturing—where hardware is constrained, lighting is inconsistent, and latency matters. I design pragmatic ML systems: clean data contracts, labeling pipelines that don’t crumble under scale, training/evaluation that tells the truth, and deploys that are observable and reversible. When milliseconds and watts count, I tune models and runtimes (ONNX/TensorRT) to hit target FPS on Jetson/desktop GPUs—without sacrificing accuracy.
Before startups, I worked on autonomy (perception/detection) and led data and inference pipelines. I also spent years in patents, which means I can help you move fast and protect what’s novel. I’m available for remote, part-time engagements: architecture reviews, performance tuning, productionization, and hands-on delivery.More...