I build production-grade computer vision pipelines that see, reason, and act — turning raw camera feeds into real business impact.
I sit at the intersection of machine learning research and production engineering — translating state-of-the-art models into pipelines that actually run on messy, real-world camera feeds.
Leading a stealth Proof of Concept for AI-driven theft prevention across Walmart international stores in Chile, Mexico, and Canada. I own the end-to-end strategy — from research direction and system architecture to live stakeholder demos that unlock multi-million-dollar rollouts.
The stack: YOLOv26 for detection, SAM2/SAM3 for segmentation, MMAction2 for temporal action modeling, and custom temporal-reasoning pipelines that correlate pose keypoints, tracked objects, and object-disappearance events to infer concealment behaviors on low-quality camera feeds.
See the full storyWhether you're hiring, collaborating, or just curious — I'd love to hear from you.