Maastricht Deprisk – ML Depression Risk PredictorLIVE

End-to-end ML pipeline with FastAPI backend + Next.js frontend. Containerized service runs on Azure Container Apps; live, non-clinical demo uses a tuned XGBoost model (AUROC 0.71) trained on 9,000+ participants from The Maastricht Study.

• ✅CI/CD: Actions builds → pushes to GHCR → deploys to ACA
• ✅Scale-to-zero, health checks, and Log Analytics for observability
• ✅x-api-key auth with server-side proxy in Next.js (no key in browser)
• ✅CORS allow-list; domain split ready (app/api subdomains)

Research prototype for educational purposes – not for clinical use

EU-27 Border vs Non-border Life-Expectancy Convergence (1995–2023)

Built a harmonised EU-27 NUTS-2 panel (NUTS-2016; 240 regions; 6,102 region-years) from Eurostat mortality/population data; estimated absolute β-convergence with TWFE and ran border event-study within countries. Pandemic caused a temporary σ-divergence, then re-convergence by 2023; border regions showed deeper dip and over-rebound.

• ✅Anchor-2019 population weights; region-clustered SEs
• ✅Within-country β ≈ −0.0047; segmented σ(log-LE) trend
• ✅Border event-study: deeper 2020–21 dip, +0.13–0.18y over-rebound by 2023

Social Platform – Khaterak

Co-founded and developed social app serving 10,000+ MAU. Built scalable backend with OAuth authentication and optimized performance by 40%.

2015-2017: Full-stack development & system architecture

Sleep ↔ Depression Longitudinal Analysis (Thesis)

Seven-year cohort study using Cox PH models with restricted cubic splines to capture non-linear effects of sleep duration on depression risk. Manuscript in preparation.

Cohort Insights Dashboard

Interactive Power BI dashboard visualizing cohort characteristics and ML model performance metrics for The Maastricht Study.