A collection of systems, developer tools, and music technology built to solve practical problems.
An automated pull request review platform that analyzes code differences, identifies high-risk changes, and detects missing test or documentation coverage.
Integrated the GitHub REST API with an LLM-backed analysis pipeline to review 300+ pull requests, process 10K+ lines per analysis, calculate file-level risk, and return structured insights with sub-two-second average latency.
An AI-powered music tool that transcribes melodies across Indian and Western notation, making the Indian classical music I grew up singing and listening to more accessible to a wider audience.
Transcribe Mode records a melody, detects Sa, displays sargam and Western notes, visualizes pitch movement, and exports MIDI. Practice Mode compares a musician’s phrase with a reference, scores pitch accuracy, evaluates meend and transitions, marks difficult moments, and tracks progress over time. The first version uses pretrained pitch extraction, tonic-relative conversion, dynamic time warping, and a small PyTorch ornament classifier.
A cloud-based search engine that indexed more than 100,000 web pages and delivered fast, accurate search results.
Deployed on AWS EC2 with MapReduce, Flask, SQLite, TF-IDF, PageRank, and multithreaded HTTP requests to improve indexing throughput while reducing compute usage.
A delivery routing system that finds efficient travel paths between locations across the University of Michigan campus.
Fast route-planning algorithms handle large campus datasets and reduce simulated delivery distance by up to 30% compared with brute-force approaches.