On SmartWealthAI, I built an end-to-end credit default model with explainable WOE/IV feature engineering to rank customers by default risk. I served it with FastAPI, tracked it with MLflow, and deployed it on AWS ECS Fargate using Terraform.
I also implemented PSI data-drift monitoring to flag when the credit model needs retraining, and developed a point-in-time Magic Formula stock screener with live market data and a Streamlit dashboard.
At Insight Solutions, I engineered features from an agent–client database that improved the probabilistic scoring model used to assign agents to clients. I also investigated daily production model failures through EDA and benchmarking, and presented model progress and business impact to the Product Owner.
At the University of Castilla-La Mancha, I designed and parallelized algorithms for learning Bayesian network structure on high-dimensional data, outperforming strong benchmarks by up to 8×. I built reproducible ML pipelines on a SLURM cluster with GitHub Actions CI/CD, and taught undergraduate courses in Java, algorithms, and concurrency.
