Specialist in investigative journalism and public-data analysis, focused on transparency, the environment, and human rights. I work at the intersection of data, AI, and reporting, building open-source tools for newsrooms.
📍 São Paulo, Brazil
🏢 Abraji — Coordinator of Projects
🔗 LinkedIn: linkedin.com/in/reinaldochaves
- 🏆 George Polk Award for Environmental Reporting (2022) — contributor to The Amazon, Undone (The Washington Post)
- 📰 Pulitzer Prize finalist (2023) — The Amazon, Undone (The Washington Post)
- 🤖 JournalismAI Fellow (2022) — London School of Economics (LSE)
- 📊 Sigma Awards finalist (2021) — CruzaGrafos
Scripts and methodology for analyzing water quality across Brazil using public data
- Tech Stack: Jupyter Notebook, Data Visualization
- Methodology: Web scraping + geospatial analysis
A World Bank data assistant for investigative journalists. Ask questions in Portuguese, English, or Spanish — the agent queries World Bank APIs and returns formatted tables with cited sources and methodological notes. Anti-hallucination is the core design constraint.
- Tech Stack: Python, Flask, Gemini / Ollama, World Bank WDI API
- Focus: Multilingual, source-cited data reporting
An MCP server that connects the OpenFEC API to AI assistants, letting you investigate U.S. federal campaign finance through natural conversation — no need to know the API directly
- Tech Stack: Python, FastMCP, OpenFEC API
- Focus: Campaign-finance investigation for journalists
An independent tool that automatically tracks bills in Brazil's Chamber of Deputies and Federal Senate that mention journalism-related topics. It collects proposals Monday–Saturday, filters by 16 keywords (press freedom, fake news, transparency, NGOs), deduplicates, logs only new developments to Google Sheets, and emails personalized HTML reports to subscribers
- Tech Stack: Python, Streamlit, GitHub Actions, Google Sheets, Câmara/Senado APIs
- Focus: Press-freedom legislative monitoring (serverless, zero-cost)
A CLI pipeline that cross-references any name with U.S. ties against U.S. public databases, producing a source-cited, confidence-scored PDF dossier. Optimized for Brazilian and Latin American names; works for any origin country
- Tech Stack: Python, Claude Code, OpenFEC (99 tests passing)
- Focus: Transnational investigation & verification
A structured OSINT skill for Claude, tailored to investigative journalists and grounded in the Berkeley Protocol and Bellingcat-style verification. Covers geolocation, photo verification, entity profiling, and social-media investigation — always producing an actionable lead sheet
- Tech Stack: Claude Skill, Berkeley Protocol
- Focus: Open-source intelligence with honest confidence levels
Specialties:
- 📊 Data Analysis & Visualization
- 🔍 Web Scraping & Public APIs
- 🤖 AI for Journalism (RAG, agents, MCP, LLMs)
- 🕵️ OSINT & Due Diligence
- 📰 Investigative Journalism
- 🏛️ Government Data
🔹 AI agents and generative-AI tools for investigative journalism 🔹 Dashboards and pipelines for public-data analysis 🔹 OSINT and transnational investigations from open sources 🔹 Training journalists in AI (open-source courses and materials)
- 💼 LinkedIn: reinaldochaves
- 📧 Email: reichaves@gmail.com