Projects

Full-stack blog and portfolio site with a custom split-flap header, blog management via admin API, and a scrolling timeline UI.

• Built a full-stack blog platform with a Next.js frontend and FastAPI backend, deployed on Vercel and Render.
• Implemented admin-authenticated CRUD API for blog post management with MongoDB as the data store.
• Designed an animated timeline component with scroll-driven progress effects and split-flap text animation on the landing page.

Real-time fraud detection system with automated data pipelines and interactive dashboards.

• Designed a robust MySQL schema and automated database creation/CSV ingestion via SQLAlchemy.
• Deployed the solution using a Dockerized MySQL instance to ensure consistent development and production environments.
• Implemented cyclical time encoding and weighted logic to identify high-value anomalies in real-time.

End-to-end multimodal ML pipeline combining visual, audio, and textual signals.

• Automated large-scale data collection using the YouTube Data API, implementing robust filtering, pagination, and failure recovery.
• Built an end-to-end multimodal ML pipeline to fuse visual, audio, and textual signals for downstream prediction tasks.

Deep learning research project focused on model optimization for species recognition and medical decision-making.

• Adapted SOTA models like MobileNet V4 through parameter optimization to improve species recognition accuracy.
• Constructed a Deep Learning Neural Network on PyTorch to facilitate high-stakes medical decision-making in rural regions.