Rishabh User

I’m an AI Engineer with hands-on experience building production-grade Generative AI, Computer Vision, and NLP systems. I focus on LLMs and end-to-end ML lifecycle—from data annotation to production deployment—so models deliver measurable outcomes, not just demos.

In my current role as an AI Developer, I built a production OCR pipeline using PaddleOCR and TesseractOCR to extract structured bank transaction data from PDFs/images, reducing manual data entry by 80%. I also developed a US Tax Form Extractor that achieved ~99% extraction accuracy and cut CPA processing time from hours to minutes.

I engineer practical GenAI products like a Generative AI Headshot System using Stable Diffusion, ControlNet, and the Google Gemini API, with custom LoRA fine-tuning for on-demand professional portraits. On the engineering side, I design scalable FastAPI backends with async processing, Redis caching, rate limiting, and Docker microservices for real-time ML inference, backed by MLflow and Vertex AI.

My work extends into optimization and applied research: I improved ensemble model accuracy by 15% and reduced setup time by 60% during an AI/ML internship. In projects like Railway Vision (YOLOv8 + BoT-SORT), multimodal medical report captioning (ViT + GPT-2 with RAG), and RAG chatbots (LangChain/LangGraph), I consistently aim for accuracy gains and production-ready reliability.