Senior Data Scientist based in Barcelona. I build predictive models and the data infrastructure that makes them possible. LTV forecasting, churn prediction, and end-to-end pipelines on GCP.
Featured projects
| Project | Use case | Stack | Key result |
|---|---|---|---|
| bigquery-air-quality-forecasting | Air quality forecasting + anomaly detection | LightGBM, dbt, BigQuery, Terraform, Cloud Run | 25-station ensemble, deployed to Cloud Run |
| banking-fraud-detection-pipeline | Fraud detection + expense forecasting | LightGBM, dbt, DuckDB, LangChain, BigQuery | Fraud BA=0.97, forecasting R2=0.76 |
| session-recommender-lambdarank | Session-based product recommendations | LightGBM LambdaRank, Item2Vec, dbt, DuckDB | NDCG@5 = 0.377, Hit Rate@5 = 76% |
| music-streaming-churn-prediction | Subscription churn prediction (KKBox, WSDM Cup 2018) | LightGBM, Optuna, SHAP, dbt, DuckDB | ROC-AUC 0.924 on temporal holdout |
| temporal-association-rules-multimorbidity | Clinical pattern mining from EHR data | Python, Apriori extensions, Fleiss Kappa | MSc thesis, validated with physicians |
Right now
- Deepening knowledge in causal inference and Bayesian modeling through production use cases at Madbox.
- Exploring how far AI can go in replacing or augmenting data work (and where it falls short)
📎 LinkedIn · 🌐 mponsclo.com