Systems, not notebooks.
Each project goes from raw data to something a person can use — with tests, documented decisions and stated limitations.
Diabetes Readmission
A famous, irregular clinical dataset: rare target, weak signal, 97% missing in one column, 716 diagnosis codes and repeated patients. How far does a well-specified logistic regression go against gradient boosting — measured with every metric that matters?
02EduData
Ask questions in Portuguese about eight editions of Brazil's higher-education statistics. Numbers come from the database — never from a model.
03Credit Default API
Estimate the probability that a credit-card client defaults next month, choose the decision threshold on purpose, and serve it as a validated REST API.
04Adaptive Recommender
A weekend playing with recommender systems: five different models, an engine that decides how much to listen to each one for each customer, and an evaluation that does not hide when the fancy idea only ties the simple one.