Sarmad G.

I build and fix AI agents, MCP servers and LLM-powered backends in Python, and take them from prototype to production. 7+ years as a backend engineer, 70+ Upwork jobs, $40K+ earned. An AI prototype that answers prompts is one thing. A production application that can safely take actions, recover from failures and serve real customers is another. I help startups and engineering teams bridge that gap. WHAT I CAN HELP YOU WITH - AI Agents & LangGraph: Tool calling, workflow orchestration, persistent state, human approvals, retries and failure handling - MCP Servers & Integrations: Connecting AI applications to APIs, databases and business systems through structured tools and secure access - RAG & LLM Applications: OpenAI and Claude integrations, retrieval, structured outputs, validation and evaluation - AI-Powered SaaS Backends: Python, FastAPI, Django, PostgreSQL, authentication, multi-tenancy, background jobs and webhooks - Production Readiness: Debugging existing AI applications, automated testing, Docker, CI/CD, cloud deployment, monitoring and performance RELEVANT WORK - LetMeKnow.ai: Built an AI-powered reminder and scheduling application using LangGraph to interpret natural-language requests and manage reminder workflows. - Hacker News MCP Server: Developed an open-source MCP server in Rust exposing six structured tools through stdio and Streamable HTTP. - OpenAI API Integration: Worked on a substantial long-term Upwork engagement integrating AI capabilities into an application. HOW I WORK I don't recommend rebuilding an application just because its AI components are unreliable. I start by reviewing the existing architecture, identifying the critical problems and agreeing a practical plan. I focus on the parts that decide whether software works in production: clear API contracts, database design, permissions, error handling, testing, observability and deployment. LET'S DISCUSS YOUR PROJECT Send me a short description of what you've built, your current stack and the biggest thing stopping you from launching or scaling. I'll suggest a sensible next step.