Jay V.

I build AI systems that run in production, not demos: agentic workflows and RAG systems that automate real business processes, and computer vision that works on real-world footage. Top Rated, 100% Job Success, 9+ years in AI, ex-Qualcomm. Most AI projects don't fail at the model. They fail on real data and real workflows: messy documents, edge cases, steps nobody planned for. That's the part I'm good at. AGENTIC AI & LLM SYSTEMS - Agentic workflow automation for a port & logistics company: disbursement and port-call workflows handling ~1,000 documents a day (invoices, SOFs, tariffs, agreements) from email, with OCR + LLM extraction, user-level access control, and RAG search across all documents. - Published research on LLM reasoning (CoreThink, arXiv), so I evaluate agents and LLM outputs with real metrics, not by eyeballing. - Agents and tool-calling workflows (LangChain, LangGraph), RAG over vector databases, document AI, FastAPI services. COMPUTER VISION - PPE compliance monitoring: custom detection models on real site footage, running on NVIDIA Jetson at ~30 FPS. - Custom image classifier: 97.93% accuracy on a diverse test set, outperforming a YOLO baseline; used in PPE and mall analytics systems. - Computer vision tracking for a UK waste-tech company, 571 hours: "took our project from an idea all the way to production, many times over." - Text reading in hard conditions: tyre sidewalls, machine nameplates, licence plates across multiple cameras. - Sports analytics: player and ball tracking with court mapping for a pickleball engine. PROJECTS I TAKE ON - AI agents and workflow automation - RAG systems and AI assistants over your documents - OCR and document data extraction - Object detection, segmentation, tracking and CCTV analytics - Edge deployment and optimisation (Jetson, TensorRT, ONNX) BACKGROUND Qualcomm, Softnautics (Senior Embedded AI Engineer), Matrix Comsec (Senior AI Engineer). HOW I WORK Clear scope and acceptance criteria before starting. Results measured on your real data. Honest feedback when something won't work. When a project needs more hands, I bring in my own team of AI engineers, web developers and annotators. Tell me what you're trying to solve, and I'll reply with how I'd approach it.