Over the past year I have worked on exciting features in the LLM product space to create a better math AI tutor tool, that is used by 450k students around the world. Architected end-to-end 'threads' feature enabling contextual mini-chats within AI responses. Led critical infrastructure migrations including Next.js API routes adoption, Vercel AI SDK v4→v5 upgrade, and tRPC procedures migration for business logic. Implemented Redis-backed resumable SSE streams for AI video generation feature, enabling reliable long-running operations with automatic recovery.

Built a scalable appointment management and scheduling system for Senegal's public healthcare starting with a regional hospital pilot. Engineered bitmap-based appointment storage solution optimized for 2G networks, reducing data payload by 80% (from 5MB to 1MB) and API latency by 40%. Designed intuitive single-action Flutter interfaces in Figma for healthcare workers with limited tech literacy.

Developed comprehensive music release management platform integrating multiple external music data APIs to create unified artist datasets using Next.js, PostgreSQL, and Drizzle ORM. Built end-to-end data scraping pipelines that aggregated and filtered music metadata from multiple sources into standardized SQL schema. Implemented white-label social media integrations with Metricool for automated artist release announcements across platforms. Developed AI-powered asset generation system for albums and tracks to enable artist inspiration.

Developed AI-powered educational features for AskCurie platform enhancing interactive learning for children. Built Curiosity Labs Studio: a content management system using Next.js, streamlining AI-generated content review workflow and reducing review time by 60%. Developed Firebase-powered weekly-reset leaderboard with GPT-4 anonymous username generation, driving 40% increase in weekly active users. Created high-performance story sharing page using Deno and Fresh framework achieving 95+ Lighthouse score alongside using open graph protocol.

Building a Scheme interpreter in JavaScript from scratch to deeply understand programming language design and functional programming paradigms. Implemented custom lexer, parser, and evaluator for core Scheme features. Currently exploring functional programming techniques to manage growing complexity and improve code maintainability. Project serves as practical exploration of language implementation, abstract syntax trees, and interpreter design patterns.