Skip to content
Henry Rincón

Brochure

Chat with my résumé

A chat that answers for my résumé — and degrades without hiding it.

Ask it about my experience in Spanish or English and it answers using only my real content, citing the source and linking it. If the AI provider is unavailable, it falls back to local search over the same index: the chat keeps working.

In production
  • 6

    switchable providers

  • 0

    persisted answers

What it does

Answers grounded in the source

RAG with a build-time index from the real content; every answer cites the section or case study and links to it. No hallucinating experience that doesn't exist.

Env-switchable LLM provider

Groq, Gemini, Azure AI Foundry, Claude or an OpenAI-compatible self-host: switching provider is switching an environment variable, zero code. AI engineering shown live.

Never down: local fallback

A circuit breaker degrades to instant client-side search when the provider fails. The first time it happened was a real credential failure — and it worked.

Guardrails and budget

The model only ever sees fragments of my own content: it has nowhere else to draw from. Questions that do not even resemble it are stopped earlier, without spending a token. Rate limit per IP, token ceiling and kill-switch. Nothing the model produces is stored.

Stack

  • Vercel AI SDK
  • Groq
  • MiniSearch
  • Zod
  • TypeScript