Projects

Published PyPI package - A Stardew Valley X Halloween-inspired terminal companion that lives in your git repository. Features pixel art companion that evolves based on coding habits, AI-generated stories, and automatic feeding through git hooks.

Production-ready AI medical education chatbot with FastAPI backend, OpenAI GPT-4o Mini integration, JWT authentication, and MongoDB database. Features LangGraph for conversation flow management and deployed on Render with scalable architecture.

Research project developing permutation-equivariant neural networks for multi-object tracking using MOT17 dataset. Implementing rotation-equivariant CNNs with e2cnn library for geometric deep learning applications.

AURORA: A region-aware Gaussian Process approximation method that identifies high-importance regions through uncertainty and density analysis. Achieves 41% average improvement in Expected Calibration Error (ECE) with 3-5x speedup over full exact GP inference. Collaboration with Alekhya Bulusu and Ming Ma.

Real-world multi-agent application built on top of the Model Context Protocol (MCP) for the MCP 1st Birthday Hackathon. Addresses slow, resource-intensive recruitment processes by automating CV screening, voice interviews, email communications, and calendar scheduling. Features specialized sub-agents (CV Screening, Voice Screening, Gmail, GCalendar) orchestrated by a Supervisor Agent, with context isolation, adaptive context compaction, and secure MCP server integration for Gmail and Google Calendar APIs.

AI-powered climate education platform built for global hAIckathon. Master climate science with AI-powered learning, interactive quizzes, and data-driven insights. Features instant answers, engaging quizzes, and real climate facts to build knowledge and make informed decisions for our planet.

Comprehensive AI-powered review system for hackathon organizers to automatically filter and evaluate submissions using multi-agent AI technology powered by CrewAI.

Drug effectiveness prediction from patient reviews using NLP and Machine Learning. Analyzes over 215,000 patient reviews to extract actionable insights for informed medication decisions.

AI-powered fitness form analysis system. Created for Cloud Run Hackathon 2025 - AI Agents Category. Provides real-time form feedback to help home fitness enthusiasts exercise safely and effectively.

Comprehensive statistical analysis of Walmart sales data with an interactive Streamlit dashboard for visualization and exploration. Developed and analyzed by Srikar MK & Alekhya Bulusu.

BrewAutomate empowers business owners with AI-driven RPA, automating finances, stock, and workflows, tracks product market & pricing. It saves your time, cut errors, boost growth & brew your business.

Comprehensive Python-based analysis toolkit for exploring and analyzing S&P 500 companies' financial data. Analyzes 502 companies across 2,012 data points, providing detailed insights into financial performance, market trends, and company metrics through interactive Jupyter notebooks and visualizations.

Engineered intelligent assessment generation system using LLaMA-based models integrated with DuckDB, processing 500+ educational documents with parameterized RAG framework.

Architected scalable food delivery platform using MongoDB geospatial indexing, enabling location-based restaurant discovery for 1,000+ registered users.

Created computer vision platform integrated with Gemini API for measuring brand visibility during sporting events, providing quantifiable sponsorship ROI metrics.

AI-powered tool that extracts and summarizes YouTube video transcripts using natural language processing techniques.

Keep track of the valuable videos you wanna watch later, paste the id into the box and add it up! A simple and efficient way to save and manage your YouTube video collection.

A Decentralised Ed-Tech Platform with Learn to Earn policy.

End-to-end development of AI marketplace platform serving 10,000+ monthly users. Analyzed usage patterns from 1,900+ AI tools through statistical modeling, reducing user search time by 30%. Built with React frontend, FastAPI backend, AWS infrastructure, and SQL database.