Mohamed F.

I design and build autonomous AI agents and LLM-powered automation systems that execute real business workflows — not just chatbots that answer questions. Over the past 6 years I've worked across software engineering, technical product management, and B2B/B2C SaaS operations. For the last year I've focused 100% on Agentic AI engineering: building multi-agent systems with LangGraph (ReAct, Plan-and-Execute, Supervisor-Worker orchestration), integrating LLMs (OpenAI, Anthropic, Groq, Gemini) with tool calling and structured outputs, and shipping production-grade automation with proper error handling, state persistence, and human-in-the-loop approval flows. What I bring that most AI developers don't: • Product thinking — as a former Technical Product Manager for B2B/B2C SaaS platforms, I understand system architecture, API design, and how to translate a business problem into a technical spec before writing a single line of code. • Reliability-first engineering — SQLite-backed state management, deduplication, exponential backoff retry logic (Tenacity), Pydantic schema validation, and fail-closed designs so your automation doesn't silently break in production. • Real integrations, not demos — Telegram bots, Google Sheets, IMAP, CRM data pipelines, Zapier/Make, MCP servers, RAG with Pinecone/Chroma. Recent work includes an autonomous multi-platform job-discovery agent (LangGraph + Telegram HITL), a production lead-generation and outbound engine for a B2B2C marketplace, and a competitive intelligence agent using RAG over unstructured market data. If you need an AI agent or automation system that actually runs reliably in production — not just a proof of concept — let's talk about your workflow and I'll show you exactly how I'd architect it. Github: string-MFathi