My Projects

Search Engine Project

• Led a 6-person team for the development of a B2B real-time search engine application with full-text NLP-based query matching and semantic relevance via embedding similarity, using agile methodologies.
• Implemented user feedback loops (ratings, citations) and AI-generated summaries, integrating React components with Neo4j, resulting in improved data tagging for the RAG pipeline.
• Scaled to 10,000+ papers and 500+ daily searches, <1 second query latency with microservice architecture, Neo4j indexing, and Docker/AWS back-end, Vercel front-end deployment.

Student Response Classifier

Led a 4-person team to train Naive Bayes, Random Forest, and Neural Network classifiers using TF-IDF, achieving 95% validation accuracy using 500 trees.

KVStore Database

A Tiered Log Structured Merge (LSM) Tree implementation in Zig for write optimized workloads using a ZipTree/SkipList Memtable. Inspired by Cassandra, Tigerbeetle, and RocksDB.

Archive Management Tool

A custom archive management tool (“kar”) from scratch, capable of creating and extracting archives with recursive directory support, modeled after UNIX tar utility. Implemented efficient memory management with Valgrind to eliminate leaks, handling multiple linked list directory data structures.

FoodBar

• Co-led a 7-person team to develop a social media app for cooking enthusiasts, enabling users to post, search, and share recipes with features like ingredient filtering, serving size adjustment, and recipe saving in Java.
• Followed SOLID design principles and adhered to Clean Architecture. Implemented user-driven functionalities, including user registration, interactive recipe reviews, and dynamic content filtering by cuisine or popularity.

MIPS Breakout Assembly

Used MIPS Assembly to implement a working version of the game Breakout.