CS thesis: multi-tier full-stack SaaS that ranks forex pairs by 9-indicator macroeconomic consensus, augmented with a Python GBDT model. Role + tier access, single-tab enforcement, multi-format report export.

Random Forest classifier hits F1=0.85, AUC=0.958 on 119k real bookings. Full ML lifecycle: EDA → feature engineering → model comparison → SHAP explainability → Gradio demo on Hugging Face Spaces.

Daily ETL pipeline: CoinGecko API → DuckDB → dbt models → Streamlit dashboard. Runs every 6 hours via GitHub Actions cron. dbt docs site published on GitHub Pages.

Track job applications end-to-end with magic-link auth, a Postgres backend, and an HTMX-driven UI. Deployed on Render with GitHub Actions CI.

URL shortener with Redis caching and a Postgres store, fully containerized and deployed on Railway with a GitHub Actions pipeline.