At XenReality, I build Cafe-Edge, an edge-deployed computer vision platform for restaurant and cafe operations running on Raspberry Pi 5.
I trained YOLOX-S detection on a 15,000+ image dataset, reaching 93% mAP, and deployed ByteTrack-based multi-object tracking without cloud GPU dependency.
I designed six KPI modules for cleanliness, security, store hours, delivery bays, and POS integration across 10+ live sites. I also built a Vite setup wizard that reduced new-site onboarding from days to under an hour.
I research practical edge AI systems, including face recognition and multi-camera retail tracking, and published work on GNN-SAC for AI-driven grid topology optimization.
