I'm an AI engineer who builds AI systems that do real work inside a business: agents that act on your tools, MCP servers that connect Claude to your data, and document and data pipelines. I test them against real data before anyone relies on them, and I tell you plainly what doesn't work yet.
I've freelanced since 2022 (Top Rated, 100% Job Success on Upwork). Since October 2025 I've also worked as an AI engineer at AY Automate, embedded with two European clients. Some recent results:
• A multi-tenant document platform with MCP tools and OAuth for an AI company. I wrote 37 isolation tests. They caught a cross-tenant data leak before launch, and all 37 pass now.
• Purchasing automation for a steel fabricator: 95-100% line match against what the planners actually ordered, across three projects.
• An email triage tool for a facilities company: 94% accuracy unattended and 97% with a person reviewing, on 87 labelled emails.
• A competitor promotion pipeline for a retail chain: about 2,100 offers from four competitors, with 98.9% of the client's own offers placed correctly in its product taxonomy.
• A pre-send QA agent for an email marketing agency. It checks campaign drafts against the approved design and catches mistakes people missed.
Tools I use: Claude Code and the Claude API, MCP, TypeScript, Python, Next.js, FastAPI, Supabase/Postgres, Google Cloud, Vercel, Docker, n8n, RAG and LLM evaluation.
I trained at A2SV (Africa to Silicon Valley), a one-year intensive in data structures and algorithms.
I'm based in Addis Ababa (UTC+3), so my working day overlaps EU hours and US mornings. Available as a contractor for US and EU teams, or full-time. I write clear English and keep you updated as I go. Send me the problem and I'll tell you honestly whether AI is the right fix.