Shivam S.

AI Agent Developer: I build production AI agents that plan, call tools and act on your data. Senior AI Engineer at EPAM, 7+ years. Top Rated, 100% Job Success. Most AI agents work in a demo and fail in production. I build them to run reliably, from a single assistant to multi-agent systems: clear tool design, guardrails, evaluation against real user questions, and secure access to your systems. AI AGENT PROJECTS I'VE DELIVERED → Multi-tool AI agent for security investigations: a central orchestrator that plans and routes requests across ~30 purpose-built MCP tools, with permission-aware API access, read-only SQL guardrails, OpenSearch, conversation memory and streaming answers. Evaluated against several hundred real analyst questions. → Text-to-SQL agent: fine-tuned Llama models on 2,500 execution-validated question/SQL pairs, reaching 90%+ query accuracy and replacing a costly GPT setup. → RAG knowledge assistants: hybrid BM25 + vector search (Milvus), cross-encoder reranking and citation-grounded answers. → Document AI: LLM extraction of invoices, purchase orders and tax forms into structured JSON. WHAT I BUILD → AI agents and agentic workflows (LangGraph, LangChain, MCP tools) → Multi-agent systems with an orchestrator, so new agents plug in without a rewrite → AI integration: connect LLM agents to your APIs, databases, CRM and internal tools → RAG chatbots and enterprise knowledge assistants → Text-to-SQL and natural-language data assistants → LLM fine-tuning and self-hosted models when data must stay private → Production AI APIs with Python and FastAPI, deployed with Docker, Kubernetes and AWS HOW I WORK → Start from your workflow and the questions users actually ask → Code computes numbers and takes actions; the LLM reasons and explains, so results stay accurate → Security first: least-privilege access, guardrails and prompt-injection protection → Evaluation set before launch, monitoring after → Weekly updates, clean code and handover documentation Tech: Python, FastAPI, LangGraph, LangChain, MCP, OpenAI and open-source LLMs (Llama), RAG, Milvus, OpenSearch, PostgreSQL, Docker, Kubernetes, AWS, Azure. Send me the workflow you want to automate and I'll reply with a short agent architecture and timeline.