Back to portfolio
Project / AI assistant

AI Portfolio Assistant

A retrieval-augmented chatbot that answers questions about my work using vector search over portfolio content.

How it works
  1. You ask a question through the chat interface.
  2. The question is converted to a vector embedding with text-embedding-3-small.
  3. Cosine similarity finds the most relevant chunks in the curated knowledge base.
  4. GPT-4o-mini generates a grounded response from the retrieved context.
  5. The response streams back through the Vercel AI SDK.
Technical details

Stack

Next.js 16, React 19, TypeScript, Tailwind CSS, OpenAI GPT-4o-mini, text-embedding-3-small, Vercel AI SDK, and custom cosine similarity retrieval.

Cost profile

The implementation uses a small model and small embedding model to keep per-conversation cost low while preserving enough quality for recruiter-style questions.

Try it

Use the chat button in the bottom-right corner, or return to the homepage for the full inline assistant.

AI-powered. Responses may not reflect exact words or opinions.