Abhishek Garg

I’m an NLP and Agentic AI Engineer with production experience in deep learning, retrieval-augmented generation (RAG), and end-to-end NLP pipeline development. I build autonomous AI systems that combine LLM reasoning with tool use via the Model Context Protocol (MCP), text-to-SQL agents, and conversational memory.

At Kinben Innovation, I automated anomaly detection by implementing and deploying UNET and WNET deep learning models, cutting manual annotation and saving an estimated 15+ hours per week. I also re-engineered GPU-based extraction logic to reduce data stitching time by 25%, and built 10+ RESTful API endpoints with Node.js and Express—while delivering a production ML backend on Google Cloud with 99.9% uptime.

On my side projects, I built a production-grade MCP database agent that exposes PostgreSQL/SQLite as natural-language endpoints, including a 6-stage SQL safety pipeline and self-correcting text-to-SQL loop that recovered 55% of initially-failed queries within 3 retries. I’ve also delivered a full-stack multi-user Enterprise RAG assistant with hybrid vector search (dense + BM25 fused via RRF), semantic chunking, and real-time token streaming, all backed by multi-tenant security and persistent chat history.