3-phase pipeline (plan → parallel search → analysis) that autonomously researches any topic. Uses FastMCP to orchestrate Claude Sonnet and Gemini Flash as specialized sub-agents, with asyncio.gather() parallelizing search across sources before a final synthesis pass.

Full-stack app built across 4 containerized services: a React frontend, Node.js/Express API, MongoDB store, and two Python/Flask backends. Supports dual recommendation modes — TF-IDF cosine similarity for fast retrieval and a pydantic-ai multi-agent pipeline for nuanced natural-language queries.

Document retrieval system using Weaviate as a vector database with hybrid search (dense + sparse). Built a RAG pipeline with chunked embeddings that lets users query a large document corpus in natural language and get grounded, cited answers.

Android app for navigating course registration and connecting with classmates. Features real-time chat, course search, and user profiles. Built following an agile process with sprint backlogs, daily scrums, and formal user/system requirements documentation.

Client-server system simulating real-time order dispatching and delivery. Concurrent driver coordination via thread synchronization dispatches orders from a centralized HQ server. Integrated the Yelp API for real-time restaurant data and shortest-path routing, with deadlock prevention and concurrency control throughout.

Web app for roommate coordination built with ReactJS and Spring Boot. Designed UI mockups in Figma and Photoshop, implemented the frontend, and wired backend connectivity as part of a 7-person team.

Retail system built in C++ with product management, user authentication, and shopping cart functionality.

Researched the impact of varying training data sizes and network configurations on GAN performance using TensorFlow. Authored a research report on the Effect of Training Data and Network Configurations on the Performance of Generative Adversarial Networks.