Pedro Rebouças Veloso
Data Scientist
- Location
- João Pessoa, Brazil
- Currently
- Data intern @ TJPB · Researcher @ TAIL
- Studying
- Data Science — UFPB
- Organizer
- Careers in Data
About
Data intern at the Tribunal de Justiça da Paraíba, working on data modeling and warehouse design for judicial statistics, and on predictive models for court performance indicators.
At TAIL — the largest academic AI league in Northern/Northeastern Brazil — I work on computer vision, currently building a system for the Paraíba State Basketball Federation's championships.
enjoying the ride :)
Work
Biu — WhatsApp Employment Assistant
Conversational assistant that builds PDF resumes from WhatsApp voice messages for people with low literacy — no forms, no typing. Also matches job listings to their profile. Team project; my role was AI engineering and research. Built with Python, Gemini API, and WhatsApp Cloud API.
Herança Tática
Web app comparing the tactical DNA of national teams across World Cups from 1930 to 2022. Historical football data is fragmented — each era has different detail levels. Similarity is calculated only over features present in both sources, with explicit confidence levels: high (6+ features, both StatsBomb), medium (3–5, mixed sources), low (1–2, basic historical data). Up to 12 tactical metrics when data allows. Built with React, Vite, Tailwind, Python, and FastAPI.
Fake News Detector
Classifier achieving 0.97 F1-score on both REAL and FAKE classes over the WELFake dataset (72,000+ labeled articles). Combines classical NLP, Transformers, and linguistic analysis. Full pipeline from data cleaning to sentence embeddings, built with Python, scikit-learn, and HuggingFace.