A working portfolio of LLM workflows, RAG systems, agent tooling and classical ML — shipped as MVPs, documented with rebuild notes, and measured against named test cases. No screenshots-as-product; the demos are live.
Every project ships with a live URL, screenshots from real runs, and a written evaluation. No "improved efficiency" claims without numbers attached.
Notes list the secrets, endpoints, gotchas, and the exact order of operations. If you can't follow them to a working copy, they get rewritten.
Working MVP, In Progress, Planned — each one means what it says. Roadmap items are not pretended to be finished.