Vaidik Garg
Student developer · Java server work & full-stack web
I'm a student who builds full-stack web projects and sends fixes to open source projects. 4 merged upstream PRs so far, in Apache HugeGraph and Apache Fineract.
On Apache HugeGraph I work on the store layer: giving RocksDB's BackendColumn a hash code that matches its byte-based equals(), so intersection and union iterators stop treating identical rows as different, and stopping ProjectionStage from clearing a column that was still held as a dedup key, which let duplicate rows through on id-only scans. I reported that second bug as an issue before fixing it.
On the Apache Fineract back-office UI (Angular) I fixed client records that showed date-only fields as midnight timestamps in the viewer's time zone.
Outside upstream work I'm building Study Vault, a course notes hub with generated quizzes and flashcards, on FastAPI and SQLite with a Tailwind front end.
- [Bug] Store limit dedup emits duplicates on id-only scans because ProjectionStage mutates the stored column in
apache/hugegraph(closed)
- study-vault: 🎓 Campus Course Notes Hub & AI-powered Quiz and 3D Flashcard Study Engine built with FastAPI, SQLite, and Tailwind CSS.
Python
The lists above are rebuilt from the GitHub API every six hours by a small script.

