Projects

Launch-Bound and Substitutable: Why Three MoE Inference Optimizations Don't Pay

Launch-Bound and Substitutable: Why Three MoE Inference Optimizations Don't Pay

An empirical study showing why three common MoE inference optimizations — quantization, custom Triton kernels, and torch.compile — fail to pay off across OLMoE, DeepSeek-V2-Lite, and Qwen3.

Majka - Postpartum Companion

Majka

Majka is a full-stack companion app that helps new mothers rebuild strength, track intake, and receive AI-assisted guidance.

SongMoodDetection

Song Mood Detection using Spotify Dataset

This project uses audio features to determine the mood of songs.

SplitSmart

SplitSmart

An application that helps you track expenses through natural language.

AIEvaluationTool

AIEvaluationTool

AIEvaluationTool is a comprehensive, end-to-end framework designed to automate the evaluation of Indian Conversational AI systems.

InstructionLLM-QnA

InstructionLLM-QnA

This project aims to build a task-specific LLM for Q&A-based tasks.

B.Tech. Thesis: ML & DL for Vapour Liquid Equilibrium Prediction

B.Tech. Thesis: ML & DL for Vapour Liquid Equilibrium Prediction

Machine Learning and Deep Learning on Vapour Liquid Equilibrium Prediction of Carbon Dioxide over Amines.

NITRKL Syllabus Bot

NITRKL Syllabus Bot

A chatbot project based on RASA and Python that showcases the syllabus and recommended books for various subjects taught at NITR.

Detecting the best waterfall spot in Tamil Nadu to set up a restaurant

Detecting the Best Waterfall Spot in Tamil Nadu to Set Up a Restaurant

This project helps a business owner or restaurant chain find the most suitable location to set up a new restaurant or branch near waterfalls in Tamil Nadu.

Exploratory Data Analysis on Student Performance

Exploratory Data Analysis on Student Performance

This project analyzes student performance across factors such as gender, ethnicity, parental level of education, lunch, and test preparation course.

Price Estimation of Used Cars

Price Estimation of Used Cars

Predicts used-car prices based on attributes such as location and vehicle age using basic ML models.