Developed an explainable machine learning framework to predict soccer match outcomes using match statistics and player performance features. Integrated SHAP-based model explanations to surface the most influential factors behind predictions, enabling analysts to interpret results and validate model reasoning. Delivered visual insights alongside prediction confidence to support tactical decision‑making and post‑match review.

Developed a harmonization framework to align heterogeneous ICU flowsheet variables across MIMIC-IV and eICU collaborative research datasets. Designed embedding-based methods for semantic matching of clinical measurements, reducing manual alignment overhead. Enabled cross-institutional predictive modeling by producing harmonized variable representations suitable for mortality and length‑of‑stay tasks. Integrated clinician feedback loops to validate mappings and iteratively refine embedding strategies for higher fidelity clinical alignment.

Designed and deployed an end-to-end data pipeline that asynchronously scrapes Reddit subreddits with Playwright, streams events through Kafka, processes them in Apache Spark Structured Streaming, and stores posts and comments in MongoDB with PostgreSQL orchestrating parsing history and control logic. All core infrastructure is containerized with Docker for reproducible, scalable deployment.