Projects

ProblemMusicians lack a lightweight way to log practice sessions and see progress over time.

BuiltA full-stack practice-tracking platform on Next.js and TypeScript, with a relational PostgreSQL schema (2 entities, 1 relationship) hosted on Neon, secure Firebase auth, and full CRUD. Now developing AI-powered recording analysis for personalized practice feedback.

Result8 users, 100+ practice log entries, zero data loss.

ProblemBoxers training solo have no real-time feedback on punch form or reaction time.

BuiltA real-time computer vision boxing trainer processing motion data at 30+ FPS with OpenCV and Mediapipe, feeding a PyTorch model that classifies three punch types, plus an analytics pipeline tracking performance across sessions.

Result87% classification accuracy; analytics pipeline improved user reaction time and accuracy by 20% across 20 sessions.

ProblemMovie buyers and casual browsers had no single place to explore films, read peer reviews, and complete a purchase without leaving the site.

BuiltA Django application where users browse available movies, read and leave reviews, and purchase copies directly through the site, with dedicated Home, About, Movies, Cart, and Orders pages.

Result5 fully built pages covering the entire browse-to-purchase flow, with support for up to 10 copies per movie, review moderation via user reports, and full order history with itemized invoices.