Software engineer working where software, data, and quantitative finance meet. Currently at JGP Crédito, on the infrastructure behind financial-markets data.
Three disciplines, one thesis: use data to solve finance problems, and engineer the systems that make the solution dependable.
Typed, tested systems — APIs, services, and the infrastructure that keeps them dependable in production.
Pipelines and models built on data that spans years, with validation that catches leakage before production does.
Financial-markets domain knowledge — the market structure and instruments the data and systems actually serve.
Unsupervised anomaly detection over 14 years of B3 equities with an LSTM-autoencoder, built so the methodology actually holds up.
Every decision is a logged ADR (18 of them, including ideas rejected with proof — Transformers, Optuna, weight decay), guarded by 32 tests and walk-forward validation against temporal leakage.
An autonomous-robot MVP built on a clean observer-pattern core, so vision, control, and results stay decoupled.
Signals are typed — Frame, Control, Result — so each subsystem reacts to events without knowing about the others. Architecture as the feature.
A queryable terminal for 1,103 B3 stocks, REITs (FIIs) and ETFs — CVM and Yahoo data in SQLite, a Swagger-documented API, an MCP server for AI, screeners, live indicator views and Markowitz optimization down to whole shares.
A deployed full-stack vault for code and prompt snippets — JWT auth, tagging, search, favorites.
A personal library of graph representations and algorithms, built from the ground up.
A browsable catalog of undergrad-thesis (TCC) ideas — filterable cards, a detail panel per idea, and two audience-scoped views over the same dataset.
Fully customized environment, optimized for my end usages and colorschemes. Used tools: tmux, zsh, ghostty, neovim.
Software Engineering Intern at JGP Crédito — building infrastructure for financial-markets data.
