I help businesses turn scattered data and manual workflows into reliable AI systems.
Cairo-based, working with Egyptian companies and remote teams worldwide.

• Turn documents into answers Make contracts, reports, and knowledge bases searchable through an AI assistant instead of long email threads and phone calls.
• Cut repetitive support work Let an AI layer handle common questions from customers or internal teams, so humans only deal with edge cases.
• Automate internal workflows Have agents read data, call tools and APIs, and push results into the systems you already use instead of manual copying and checking.
• Make decisions faster Pull the right information out of messy systems so managers can see what is going on without chasing ten different people.

• AI knowledge assistant for internal docs RAG based system on top of PDFs, Notion, Google Drive or internal storage that can answer questions in Arabic and English.
• Agent that runs business workflows An agent that reads data, calls APIs, writes to CRMs or ERPs, and reports back what it did, with clear logs and limits.
• Cleanup and hardening of an existing AI prototype Taking a “works on my laptop” LLM demo and turning it into a reliable, testable service with clear architecture and monitoring.

I am a software engineer with a focus on Python-based AI systems and agentic architectures. My background in embedded systems makes me treat AI like any other critical system: it should be explainable, efficient, and verifiable, not just impressive in a one-time demo.

That means I care about:

• Clear boundaries between what the AI decides and what the system enforces.
• Repeatable behaviour with proper evaluation, not just “it worked in front of the client once”.
• Simple, modular architecture that another engineer can understand and extend later.

Most of my work ends up using combinations of: LLM and RAG pipelines, agentic workflows, and Python services around them (LLMs, vector search, FastAPI, LangChain, LlamaIndex, Supabase, Redis, Docker and similar tooling). But the goal is always the same: make the system useful and predictable for the business.

• 🧠 RaftRag A RAG oriented template that shows how I structure retrieval augmented apps for real data and real users.
• 🧩 MCP_Rag_Agent An agent that mixes RAG with tool use so it can both answer questions and take actions over a knowledge base.
• 🤖 Rag_Agent Experiments in agent behaviour on top of RAG pipelines, focused on reasoning and multi step workflows.
• 🏗️ local-ai-stack A local first AI stack for running LLMs, retrieval, and automations on your own infrastructure instead of sending everything to the cloud.

You do not need to know the right buzzwords to start a conversation. If any of these are true, I can probably help:

• You have important information trapped in documents, chats, or tools that people keep asking about.
• Your team repeats the same tasks again and again that could be turned into a clear workflow.
• You already have an AI prototype, but you do not trust it enough to put it in front of customers.

👉 Email: mahmoud.bayoumi.pp@gmail.com
Tell me what your business does and what slows you down. We can figure out if AI systems can actually help, not just make a nice demo.