About me

Hey there! Welcome!

I am a Lead Machine Learning Engineer at Nubank, where I work on the foundations of AI agents: how to build, evaluate, and continuously improve customer-facing agents at 100M+ user scale. My focus is the agent lifecycle — turning ad-hoc prompt tuning into a rigorous, evaluation-driven discipline through calibrated LLM-as-a-judge, automated prompt optimization, automated harness optimization, and simulation-based testing.

A through-line runs across my career: leading both research and real-world application of machine learning. Before agentic AI, I spent years leading AI against financial crime at Nubank — risk and anti-money-laundering models and ML platforms that scaled across the bank’s global expansion.

I hold a Ph.D. in Computer Science and Computational Mathematics from the University of São Paulo, advised by André C. P. L. F. de Carvalho. The thesis, Automating Machine Learning Pipeline Design via Metalearning, was recognized as the 2nd best Computer Science Ph.D. thesis of the year by the Brazilian Computer Society (CTD 2026). That research produced the open-source pymfe library and a 2nd-place finish in the NeurIPS 2021 MetaDL challenge, with works published at NeurIPS, KDD, JMLR, and Artificial Intelligence Review.

My research interests include machine learning, AutoML, metalearning, AI agents, and evaluation-driven ML systems.