b.peres

Bárbara M. Peres · Full-Stack Developer & AI Engineer, Portugal

The interactive version of this page did not load. Reloading usually fixes it. Otherwise the links below still reach me.

Civic Platform · 2026

SNS Memória Digital

20 years of the Portuguese Health Service, verifiable on Arquivo.pt

Team
Solo
My role
Builder & Researcher
Stack
Python · Flask · React · RAG · Groq Llama 3.3

Visit snsmemoria.pt

  • 51 Commitments verified
  • 35+ Arquivo.pt captures
  • 4 Arquivo.pt APIs

A civic platform about 20 years of the Portuguese National Health Service (SNS, 2005–2026). Each claim sits next to its primary source, and the platform references 35+ verified Arquivo.pt captures. There are 51 government commitments classified by fulfilment, an interactive health-by-district map, a government comparator, and an AI agent that answers questions with RAG over the dataset and a live crawl of Arquivo.pt.

Python · React · Vite · Groq Llama 3.3 · RAG · Arquivo.pt API · Render · Civic Tech

The problem

Public-policy debate about the SNS runs on selective memory and on claims nobody checks. Citizens need one place to check what each government actually promised, what was delivered, and which legal acts back the facts. Each claim on the site sits next to its primary source.

What I did

I designed and built the backend in Python (Flask + Gunicorn), with a React + Vite frontend on Render. I curated 51 health commitments across 7 Constitutional Governments from official sources: Diário da República for the decree-laws, historico.portugal.gov.pt for government composition, ACSS for family-doctor coverage, INE for district demographics. I verified 35+ Wayback URLs to Arquivo.pt, all returning HTTP 200. I wired in four Arquivo.pt APIs: TextSearch for the live RAG crawl, Wayback for the embedded captures, CDX for the institutional-site timeline, and Screenshot through a cached proxy. The AI Agent runs on Groq Llama 3.3 70B, with Llama 3.1 8B as fallback and a local-search path for when there is no key. I added a transparency section that says which numbers are read directly and which are derived: district-level family-doctor counts, for instance, come from regional ARS rates × INE district population.

The result

The platform is live and has seven sections: commitments, map, comparator, quiz, glossary, before-after, AI agent. It carries 51 verified commitments, 35+ verified Arquivo.pt captures, four Arquivo.pt APIs and a data-provenance table. Submitted to Prémio Arquivo.pt 2026.

All projects