Sibghat U.

AI engineer building production AI agents, custom MCP servers and automation for business systems — agents wired to your real tools through Model Context Protocol, LangGraph orchestration, n8n + OpenAI/Claude workflows, and Python/FastAPI backends, with the guardrails, logging and human-in-the-loop approvals that keep them running after the demo. Most automations work in the demo and break in month two. I build the opposite: scoped flows, error handling and retries, audit logs, and documentation your team can run without me. 🔌 MCP SERVERS — MY FLAGSHIP SERVICE I build custom MCP servers that let Claude, ChatGPT, Cursor and Copilot work directly with your systems: your REST APIs, databases (PostgreSQL, Oracle, MySQL), SaaS tools and internal platforms. ✅ Tool design, auth, safe write-paths with approval steps ✅ Docker deployment + setup in Claude Desktop/Enterprise, Claude Code or your own app ✅ Shipped: 23 AI developer skills with MCP integrations (Azure DevOps, PostgreSQL, Playwright) that take a work item from plan to pull request — in daily use at Pakistan's largest telecom 🤖 AGENT ORCHESTRATION, FRAMEWORKS & MODELS ✅ LangChain & LangGraph: planning, tool calling, state and memory management, structured outputs, retries, human-in-the-loop checkpoints ✅ Multi-agent systems with CrewAI and AutoGen; LlamaIndex for document pipelines ✅ Open-source & private large language models (LLMs): self-hosted with Ollama and Hugging Face (Llama, Mistral, Qwen) when your data cannot leave your infrastructure ✅ Fine-tuning & evaluation: LoRA/QLoRA fine-tuning of open-weight models, LLM evals and regression testing of prompts before they ship ⚡ AI AUTOMATION & AGENTIC WORKFLOWS ✅ n8n + OpenAI/Claude pipelines with webhooks, API integration, approval steps and scheduled runs — orchestrated agentic workflows, not single-prompt scripts 🛠️ LLM APP BACKENDS & APIs ✅ Python APIs with FastAPI, Django and Flask: REST endpoints, background jobs, webhooks, OAuth/JWT auth ✅ RAG (Retrieval Augmented Generation) with vector search, Deepgram speech-to-text, document parsing ✅ Data for AI: pandas and NumPy processing, high-volume ETL (170 GB streaming XML → analytics-ready CSV), PostgreSQL 🚀 SHIPPED RESULTS ✔ 23 MCP-integrated AI developer skills automating plan-to-pull-request flows ✔ n8n + OpenAI automated workflows that score inbound jobs, draft proposals and send weekly reports ✔ Call-recording transcription pipeline on AWS S3 + Deepgram for a UK client ✔ AI-assisted approval workflows inside a production ERP — value-based rules enforced server-side, full audit logs ✔ Java 17 microservices on Google Cloud Run for a US healthcare platform 🏆 WHY TEAMS PICK ME 5+ years as a backend and integration engineer before AI — REST APIs, Oracle ERP, Java Spring Boot, PostgreSQL. I connect AI agents to enterprise systems most AI freelancers can't touch, and I treat reliability as a feature: retries, monitoring, rollback plans and docs on every build. 🧰 STACK MCP (servers & clients) | LangChain | LangGraph | CrewAI | AutoGen | LlamaIndex | Claude API | Claude Code | OpenAI API | Ollama | Hugging Face | n8n | RAG + vector databases (Pinecone, pgvector) | Python (FastAPI, Django, Flask, pandas, NumPy) | Deepgram | PostgreSQL | Docker | Google Cloud Run | REST APIs & webhooks 🤝 HOW I WORK Clear scope first, small deliveries you can test, documentation and a proper handover. 📩 Want AI assistants to work with your systems? Send me the API or database you want exposed — or the process you want automated — and I'll reply with a short build plan before you spend anything.