At Jio Platforms, I build production-grade computer vision, video analytics, and real-time AI systems for retail, warehouse, and internal knowledge workflows.
I built a multi-store footfall analytics pipeline using person detection, ReID, ByteTrack, TensorRT, PostgreSQL/pgvector, and Kafka, resolving accuracy issues against manual ground truth. I also improved video classification accuracy by 35% with VideoMAE and migrated YOLOv4 to YOLOv7 for a further 10% detection-accuracy gain.
I've developed an on-premise Mistral-7B RAG knowledge assistant for 500+ internal users, a VLM-based video understanding pipeline, and a unified multi-event detection system with over 80% precision. I also mentor interns on annotation, dataset curation, quality control, and labeling standards.
