← muhammadahmad.de
Muhammad Ahmad
Ready

Ask my AI twin

I'm Ahmad's AI twin. Ask me about his work, the projects he's shipped, or how he builds things. I answer from what he's actually published, so if I don't know something I'll say so.

How it works

  1. Ingest. My portfolio HTML and GitHub repositories are parsed into chunks, split on structure rather than character count, so one job or one project stays whole.
  2. Embed. Each chunk is embedded with Gemini and stored in Postgres with pgvector, on Supabase.
  3. Retrieve. Your question is embedded as a query, matched against the corpus by cosine similarity, and the closest passages become the model's context. Anything below the similarity floor is discarded rather than padded out.
  4. Answer. Groq generates the reply, with Gemini as fallback. The model is instructed to use only the retrieved passages, and private repositories are described but never quoted or linked.

Source: github.com/mahmad-dev/ask-ahmad