Software engineer and advanced Data Science student at UNSAM, in Buenos Aires. I work mostly on the backend, and I like problems where I have to learn something new to solve them. My interests right now are event-driven services, data pipelines and warehouses, and machine learning projects.
- tickwright is a Python event-driven trading engine. It has a crash-safe order lifecycle, idempotent recovery and exchange reconciliation. You bring the strategy. It handles the rest.
- polyglot-dwh (contributor) is a data warehouse. An operational PostgreSQL schema feeds a star schema warehouse through a daily Python ETL on GitHub Actions. The warehouse exposes SQL data mining functions and pgvector semantic search.
- potrerillos-wind-forecast (contributor) predicts wind 12 hours ahead for a mountain reservoir, using 8.5 months of station data. I compared SARIMA, LassoCV and LSTM models with walk-forward evaluation. The best one cut RMSE by 28 percent against a seasonal-persistence benchmark.
- buscasam (contributor) is an academic search platform on GCP that combines semantic and full-text search. I shipped features end to end, such as document publication, a moderation queue and notifications.
- Languages: Python, SQL, R, TypeScript
- Backend and distributed: FastAPI, asyncio, Kafka, REST, WebSockets, SQLAlchemy, Alembic, Pydantic
- Data: PostgreSQL, BigQuery, Redis, QuestDB, pgvector, ETL, dimensional modeling
- ML and statistics: scikit-learn, statsmodels, TensorFlow, Keras, PyMC, pandas, NumPy, time-series forecasting, Bayesian statistics
- Infra and quality: Docker Compose, GCP, Digital Ocean, Git, GitHub Actions, pytest, hypothesis, mypy, ruff
B.Sc. in Data Science at UNSAM, around 75 percent done, expected November 2027. Relevant coursework includes machine learning, statistics and inference, time series, databases and software engineering.