Mustafa H.

If you need an AI system that actually does something, like qualifying leads, answering customer questions from your own documents, screening support tickets, or automating a manual process, instead of a basic chatbot wrapper, that's the work I do. I build agentic AI systems. Multi agent workflows that split a task into steps (retrieval, reasoning, verification) instead of asking one model to handle everything at once. Retrieval Augmented Generation (RAG) pipelines that ground answers in your own documents so the AI isn't guessing. Voice AI agents. And the backend to run all of it: Python, FastAPI, OpenAI's API, vector databases like Pinecone, and cloud deployment on GCP or AWS. My path here wasn't the usual computer science route. I spent 20 years in telecom, managing RF network deployments and technical programs for companies like AT&T, Verizon, and Meta. That work taught me how to run a project end to end: scope it, manage the moving pieces, and report on it in a way people actually understand. A few years ago I got curious about AI and started learning agentic systems seriously. That led to a role at a specialized AI agency, first as an AI Associate, then moving up to Lead AI Engineer, working alongside data science, engineering, and QA teams. Projects I worked on there include a multi agent system that uses causal ML to figure out which customers actually respond to discounts instead of blasting everyone with promo codes, a RAG based research assistant that cites its sources, a support agent for a manufacturing client with guardrails that hand off questions it can't confidently answer, and a bilingual voice assistant for banking style call center inquiries. The telecom background turned out to be more useful than I expected. Most AI freelancers can write code but haven't managed a real technical program with deadlines, stakeholders, and a client who needs plain English updates. I can do both. If you're trying to automate something repetitive, want a support or research assistant you can actually trust, or need someone who can scope an agentic AI build and see it through, send me a message with what you're working on.