Technical

  • Glasshouse (2026 – present) – Modular fact-checking toolkit, in development, covering claim detection, claim selection, verified claim retrieval, evidence retrieval, verdict prediction, and justification production. Each stage keeps an inspectable record of its inputs and outputs. Builds on VeriTrace, an observable AI fact-checking prototype developed during a 24-hour hackathon.
  • LexGPT (2026 – present) – Legal research workbench built on corpora I collect and curate from Brazilian courts and legislative sources. Supports complex legal questions through source-linked answers and inspectable evidence, alongside collaborative document editing and shared research projects.
  • News Quality – Bluesky labeler that annotates shared links with news-source quality ratings. Sourced from Lin et al. (2023).
  • Judex (2026) - Scraper for Brazilian Supreme Court case information. Python library.
  • Obras DF (2025) - Analysis of planned federal investment in infrastructure projects in Brazil’s Federal District (in Portuguese). Interactive report and Jupyter notebook.
  • Intersect (2025) - NLP-powered job board. Uses NLP (embeddings, UMAP, HDBSCAN, reranking) to find more relevant roles. Streamlit app.

Qualitative

Creative

  • Prophetic Strategies (2024 – present) – Data/art project about oracular philosophy and AI using RAG and fine-tuned LLMs. Interactive web app.
  • Learning to Love the Labyrinth (2023) - Art-house video game. Made in Godot. Selected for Indiepocalypse #44, an independent game anthology.
  • WEAV (2021) - Elementary cellular automata visualiser. Web app made in Unity.

Publications & Talks

The midnight sun: Popper and rationality without foundations (2024)
Psychoanalysis and Science Research Group at APOLa SP
Two lectures on philosophy of science

Xenofeminism (2025)
Thinking Out Loud project at the London Interdisciplinary School
Lecture on critical theory and philosophy of nature

Awards

Hackathon Winner: AI safety hackathon (Holistic AI & UCL AI Society)
Stereotype detection in large language models using sparse autoencoders

Other


For collaboration opportunities or to discuss my research, please get in touch.