Created a search engine for Wikipedia, combining PageRank and TF-IDF for relevant results. Developed a scalable MapReduce framework, similar to Hadoop, for efficient distributed processing of large datasets. Implemented fault-tolerant communication between Manager and Worker modules using sockets and concurrency. Created REST APIs with Flask for seamless interaction with search engine components.

Consulted a kinesiology professor in game-based exercise to optimize the game's design and effectiveness. Created engaging game features (in Unreal Engine); e.g. coin collecting and fighting zombies/robots. Developed features such as dynamic height adjustment, procedural generation, and modular level development. Garnered positive feedback during playtesting and showcase, indicating an enjoyable and effective fitness experience.

Created an AI-powered Chrome extension using JavaScript and LLMs to help users stay focused by intelligently blocking distracting web pages. Features include intuitive UI, customizable block/allow lists, smart caching, and more. Built a robust and scalable backend system using Python with Flask, PostgreSQL for data storage, Fly.io for auto-scaling servers, redundancy with failover to more advanced language models, and an A/B testing system for continuous improvement.

Led the creation of an automated system for detecting Formula 1 track limit violations, leveraging the Segment Anything Model (SAM) for accurate segmentation of cars and tracks from onboard camera footage. This approach is adaptable across various tracks and car models without the need for additional training. Devised and tested a novel algorithm for generating dynamic road masks and detecting violations. Validated with both simulated and real race footage. Achieved high recall, precision, and F1-scores.