Salah Eldin E.

I build AI agents, RAG chatbots, and automation that survive contact with real users. Most of my clients come to me after something else broke in production. Top Rated Plus with a 100% Job Success Score on Upwork. The systems I've shipped are still running for the people I built them for. AI Agents & Automation Custom agents and multi-agent systems in LangGraph and n8n that take over repetitive work. I build human-in-the-loop approval into them, so the AI does the heavy lifting but never acts on anything important without your sign-off. I'm currently building a multi-agent email coordinator for a US financial services company, taking triage, routing and lead handling from human-operated to agent-operated with approval gates exactly where the risk sits. Their output carries regulatory exposure, so output validation and audit trails are standard practice for me rather than an afterthought. RAG Chatbots & Knowledge Systems Assistants trained on your own documents that answer with citations instead of inventing things. I use hybrid retrieval combining vector search with keyword search, and I enforce data isolation at the database level, so in a multi-user system one client's data can never surface in another's answers. Voice AI & Integrations Sub-800ms voice agents built with client-side voice detection, streaming speech-to-text, and sentence-boundary chunking so the reply starts before the model finishes thinking. Clean API integration with OpenAI, Claude, and Gemini. I work natively in Arabic and English, including dialect handling, which matters if you're serving the MENA market. Backend that holds up Python, FastAPI, PostgreSQL, Docker on AWS, GCP or Azure. I've built pipelines ingesting around 5 million records daily with delta diffing and full audit history. How I work Most of my work happens in the parts clients never see: error handling, retries, and the architecture that keeps your app responsive while the AI thinks. That's the difference between a system you trust and one you babysit. I'm also honest about fit. If I'm not the right person for your project, I'll tell you upfront rather than take the work. You can read my code before you hire me. My GitHub is public under salah55s, with production RAG and multi-agent projects, full source and documentation. If you're adding AI to your product or replacing a slow manual process with something that runs on its own, message me with what you're working on. I'll give you my honest read before you commit to anything.