Projects

NeuroGuard

ML web app to predict stroke risk with 78.8% accuracy and 76.9% recall using SMOTE-enhanced Logistic Regression. Built with Flask API, Gemini 2.0, and a Vite-powered React frontend.

Languages: Python, Flask, scikit-learn, React, Gemini 2.0, Vite

Activity Recommender CLI

C-based CLI tool that suggests group activities using 300D GloVe embeddings and cosine similarity. Features NLP pipeline and a minimal Next.js frontend wired to a C REST API via microhttpd and cJSON.

Languages: C, GloVe, NLP, Cosine Similarity, microhttpd, Next.js

GitHub PR Summarizer

GitHub Action that generates context-aware PR summaries using Gemini 2.5 Flash. Parses diffs, titles, and file changes. Includes YML workflows and secure setup with GitHub Secrets.

Languages: GitHub Actions, Python, YML, Gemini, CI/CD

Pixiefie

Generates branded, shareable ID cards from GitHub and Reddit handles. Fetches profile pic, name, and username; embeds live QR code.

Languages: Next.js, Tailwind CSS, Vercel, QR Code, GitHub, Reddit

Private Asset CDN

Edge CDN using Cloudflare Workers and Backblaze B2 to serve assets from a private bucket without routing traffic through the main server. Uses signed B2 download tokens internally and aggressive edge/browser caching to minimize origin hits.

Languages: Cloudflare Workers, Backblaze B2, CDN, Edge Caching, Origin Shielding, TypeScript