Health answers, not more searching. | Frederico Luz

Ask a question. Understand the answer.

A medicine database can contain the right information and still leave a reader lost. They need an explanation, not another technical document to work through.

The portal’s central feature is a conversational health assistant. Readers ask in their own words and get a simple explanation with sources they can check. They can ask a follow-up when a term or an answer still isn’t clear.

HealthPortal conversation explaining contraindications in plain language, with a linked medical reference

The client’s knowledge, available through conversation

I built the chatbot around retrieval-augmented generation (RAG), so each answer can draw on relevant source material—not just the model’s general knowledge.

RAG architecture: source material is chunked and embedded into a vector index. A medical query is embedded and matched against that index. Retrieved passages, the original question and chat history form the model’s prompt. The model generates and simplifies the answer, returned with source references.

Make the articles easier to understand, too

Alongside the chatbot, I built an editorial workspace for turning technical health articles into accessible reading. Editors import a URL or paste text, compare the original and rewritten article, then refine it with an AI assistant or edit directly before approving or rejecting it.

The rewrite does the first pass; the editor keeps control of the meaning.

HealthPortal editorial workspace with an article URL, a technical original and plain-language rewrite side by side, and review controls

The start of an ongoing collaboration

I built both the reader-facing assistant and the editorial tools during my first professional contract with Hexa. I went on to build the first working version of Growth OS with their product manager, taking that collaboration into articles and social content for agencies.