Deepak Kumar Jha
AI Demos — Senior Full-Stack Engineer · Technical Lead
Three systems, live on my own infrastructure, each with a one-click demo account against its real API — no signup, nothing mocked. This is the difference between claiming production AI experience and handing you a URL where it's running.
Privacy-First RAG
LiveRetrieval-Augmented Generation on self-hosted Llama 3 and Mistral via Ollama. Built for cases where data cannot leave the client’s infrastructure — regulated finance, government, healthcare. Upload a PDF, ask questions, get answers with citations and per-passage similarity scores.
Multi-Model Routing
LiveRoutes each request between cloud LLMs (Gemini, OpenAI) and local inference based on cost, latency and data sensitivity — and shows the decision, the fallback chain and the money saved, per request and on a live dashboard.
Agentic Workflows (MCP)
LiveA Model Context Protocol agent connected to a read-only SQL database, a live weather API and a calculator — answering multi-step questions with every tool call, result and recovery from failure streamed to the screen as it happens.
A production LLM feature already shipped in a live product — see the
SmartSitting case study →