This profile is my lab notebook. My day job lives on LinkedIn; what lives here is the independent work — studies, experiments and projects I build to understand a technique end to end, from the math to the running code.
- Physics first. I came to ML from mathematical physics and quantum information, and I still approach models the way I approach a physical system: assumptions, derivation, then experiment.
- Learning in public. Many repos here are structured study tracks — unsupervised methods, deep learning fundamentals, distributed processing — written to be re-read and reused.
- Current obsession. How to measure LLM and agent systems honestly: evaluation design, LLM-as-Judge reliability, determinism and cost-aware inference.
exploring → agent evaluation harnesses · LLM-as-Judge calibration
reading → retrieval quality metrics · cost-aware inference
revisiting → quantum information ideas that map onto ML
| Project | What it explores | Stack |
|---|---|---|
| Unsupervised-ML | Clustering, factor analysis and correspondence analysis, implemented side by side in R and Python. | Python · R · scikit-learn |
| DeepLearning | Regularization, batching, optimizers, hyperparameters and cross-validation, built from first principles. | Python · PyTorch |
| ApacheSystemML | Activity prediction on distributed data — processing, preparation and modelling with PySpark. | PySpark · Spark ML |
| Embeddings | Testing the quality and robustness of categorical embeddings on a low-feature flight-price dataset. | Python · Deep Learning |
AI · ML · GENAI
LANGUAGES · DATA
CLOUD · INFRA
| Degree | Institution | Year | |
|---|---|---|---|
| 🎓 | M.Sc., Mathematical Physics — quantum information & computation | University of São Paulo · exchange year at Uppsala University, Sweden | 2023 |
| 📊 | MBA, Data Science & Analytics | University of São Paulo | 2024 |
| ⚛️ | B.Sc., Physics | Federal University of São Carlos | 2021 |
CERTIFICATIONS · verify on Credly
Português · English · Svenska