Eslam Hamed
Technical Lead & AI Solutions Architect
Building enterprise AI systems, cloud platforms, and production-ready software
Contract and fractional work
Work with me
I build production AI systems for companies that already have real data and real users: retrieval, agents, and the evaluation and operations that make them safe to ship. Ten years in software, six of them on AI platforms on Azure, and every claim on this page is backed by a case study you can read.
You have a retrieval or agent system, or a plan for one. I measure retrieval quality, cost, security and evaluation coverage, and hand back written findings with numbers, a ranked fix list, and a working reference for the top item.
One agentic feature from brief to production: evaluations first, then the tool-calling agent, observability, and the deploy pipeline. On .NET and Azure, or Python. You get the shipped feature and a runbook your team can operate without me.
For teams adding AI to an existing .NET product: architecture ownership, code review, help with hiring, and a weekly written status that carries the numbers, not adjectives.
What you get
- Numbers you can reproduce. Every figure in a handover comes with the command or the test that produced it.
- Answers with receipts. Where an AI system produces a number, the number comes from a deterministic tool and carries its source. The model plans and explains; it never computes.
- No lock-in. Your repository, your cloud, your keys. Documentation and a runbook are part of the deliverable, not an extra.
- Fixed scope, fixed price. Agreed in writing before the work starts, with the dates.
Proof
Three pieces of work, written up the same way: the context, the constraint, the decision that mattered, and the result.
- Auto-RAG and MerlinX: enterprise retrieval for 7 to 10 customers, then the agent layer on top.
- An AI analyst whose every number carries a receipt: built to impress a boardroom and pass a security review at the same time.
- Career Rafeeq: a paid AI SaaS shipped solo in 14 days.
How it goes
- A 30-minute call about the problem and the data you have. No slides.
- A written proposal within two working days: scope, price, dates, and what "done" means.
- A short written status every week while the work runs, with the numbers.
- Handover with the runbook, and a fixed window afterwards for questions.
Start the conversation
Send a few lines about the problem and where the data lives. I reply within one working day. If you would rather check fit first, ask my AI agent; it answers from my CV and the case studies.