Featured Projects
Explore a collection of my most significant projects, showcasing expertise across various domains of computer science and software engineering.
END-TO-END APPLICATION
Project Explainer
An advanced natural language processing tool that leverages large language models to generate summaries of GitHub repositories. It offers both a Python module and a UI interface, utilizing abstractive and extractive summarization techniques to provide concise project overviews.
DISTRIBUTED SYSTEMS
Consistent Hashing for Arrow Flight
Enhances Apache Arrow Flight with consistent hashing for optimized load distribution across server clusters. Implements a scalable and fault-tolerant system for efficient data retrieval and storage in high-performance analytical workloads.
FULL-STACK APPLICATION
Box Office: Theater Management System
Comprehensive web app for movie theater operations, including ticket booking, membership systems, and employee tools. Built with Next.js, Node.js, MongoDB, and AWS. Features responsive design and Stripe integration for payments.
SYSTEM ARCHITECTURE
Quantum Taskflow
A high-performance distributed task scheduler using gRPC for efficient inter-node communication. Features low latency, high throughput, scalability, and fault tolerance. Utilizes advanced scheduling algorithms and optimized protocols.
HIGH-PERFORMANCE COMPUTING
Parallel Air Quality Data Processing System
High-performance system for parsing and analyzing air quality data using OpenMP, MPI, and POSIX shared memory. Features efficient data transfer, scalable analysis, and real-time visualization. Demonstrates significant performance improvements through parallel processing and shared memory usage.
IMAGE PROCESSING
RGB To Lab Color Space Conversion
Implements a conversion system between RGB and Lab color spaces, addressing RGB limitations in perceptual uniformity and device independence. Utilizes complex mathematical transformations to enhance color representation accuracy for digital imaging and computer vision applications.
DATA SCIENCE
COVID-19 Data Analysis
Python-based project analyzing COVID-19 data, including case counts, transmission patterns, and vaccination rates. Utilizes exploratory data analysis, statistical modeling, and data visualization to identify trends and hotspots, aiding public health efforts and policy-making.
SPORTS ANALYTICS
IPL Stats Analysis
Comprehensive analysis of Indian Premier League (IPL) cricket data using Python. Explores team performance, match outcomes, and tournament trends. Investigates factors like toss impact and identifies top-performing teams and seasons, providing insights for cricket enthusiasts and demonstrating data-driven sports analysis.
MULTI-DOMAIN DATA SCIENCE
Triple Threat Data Science
Three-part project showcasing diverse data science skills: retail data EDA, unsupervised ML predictions, and supervised ML modeling. Demonstrates proficiency in data analysis, visualization, and various machine learning techniques.
WEB APPLICATION
Student Database Management System
Efficient solution for managing student information, featuring profile management, course tracking, attendance monitoring, and grading system. Streamlines administrative tasks and enhances communication in educational institutions.
CROSS-PLATFORM GAME
Tic-Tac-Toe Web and Mobile Game
An engaging Tic-Tac-Toe game implemented as both a web application and an Android APK. Features responsive design, smooth animations, and an interactive user interface, providing entertainment across multiple platforms.
DEEP LEARNING
Grayscale Image Colorization
An advanced deep learning project for automatic colorization of grayscale images using global and local features with skip connections. Utilizes transfer learning with InceptionV3 for global feature extraction, aiming to produce realistic and color-accurate results without human intervention.
CYBERSECURITY
Cryptography and Network Security Lab Suite
Implementation of cryptographic algorithms and security protocols, including classical ciphers, modern encryption standards, key exchange protocols, and digital signatures. Demonstrates practical applications of data protection and secure communication techniques.
DEEP LEARNING RESEARCH
Neural Network Architecture Evaluation
Comparative study of various neural network architectures (Le-Net, ResNet, ResNeXt, GoogleNet, DenseNet, VGG16, EfficientNet, SE-Net, LSTM, GRU, GAN, Transformers) for a common task. Evaluates accuracy to determine the most effective model.
SOFTWARE DESIGN
Java Credit Card Processing Application
A Java-based application that processes credit card records from various file formats (CSV, JSON, XML), validates card numbers, and creates appropriate credit card class instances. Implements Strategy and Factory design patterns for flexible file parsing and card type creation.
NLP
DistilBERT Fine-Tuned for Named Entity Recognition
A DistilBERT model fine-tuned for Named Entity Recognition (NER), capable of identifying and classifying entities like names, places, organizations, and dates in text. Achieves high accuracy (97.5%) and F1-score (0.950) using optimized training techniques and custom datasets.
CONVERSATIONAL AI
LLM-Powered Conversational Retrieval Chatbot
A conversational AI chatbot leveraging large language models (LLMs) to provide intelligent responses and assist users in various tasks. The chatbot is designed to understand natural language queries, generate human-like responses, and integrate with multiple platforms for enhanced user experience.
NLP OPTIMIZATION
Finetuned-Accelerate.MLM
Developed and fine-tuned a Masked Language Model (MLM) using Hugging Face Transformers and Accelerate for optimized training. The project focuses on efficient model training, leveraging parallelism and gradient accumulation to enhance speed and accuracy.
CREATIVE AI
Movie-Name-Generator.LLM
A Large Language Model (LLM)-powered generator that creates unique and contextually relevant movie names based on genre, theme, and keywords. The model is fine-tuned to produce creative and engaging titles.
AI-POWERED NEWS AGGREGATION
NewsApp.LLM
A news aggregation and summarization app powered by Large Language Models (LLMs) that fetches real-time news articles, extracts key insights, and presents concise summaries for quick consumption.
NLP LEARNING
NLP-Hugging Face Course Implementation
Completed a hands-on NLP course on Hugging Face, learning transformer architectures, pre-training, fine-tuning, and model deployment. Explored text classification, named entity recognition (NER), summarization, and translation using Hugging Face Transformers and Datasets.
GENERATIVE AI PORTFOLIO
GenAI-Projects
Developed multiple Generative AI projects, leveraging LLMs, VAEs, and GANs for text generation, image synthesis, and creative AI applications. The repository includes fine-tuned models, prompt engineering techniques, and real-world implementations of generative AI.
LLM ORCHESTRATION
LanGraph
Developed a graph-based orchestration system for LLMs, allowing complex multi-step workflows and decision trees. Implemented node-based reasoning, contextual memory, and prompt chaining to optimize AI responses and structured interactions.
AI FITNESS COACH
Squat-Workout-Recommender.AI
Developed an AI-based squat workout recommender that analyzes squat posture using computer vision and suggests optimized workout routines. Implemented pose estimation and deep learning techniques to track movements and provide real-time feedback.
VISUAL SEARCH ENGINE
Image Search System.AI
Developed an image search system leveraging CNN-based feature extraction and vector similarity search to enable text-to-image and image-to-image retrieval. Integrated pre-trained vision models and efficient indexing techniques for fast and accurate results.
CI/CD AUTOMATION
App-Github-Action
Explored GitHub Actions for automating build, test, and deployment workflows in software projects. Implemented custom workflows, job dependencies, and environment configurations to enhance development efficiency.
HACKATHON PROJECT
CloudSprint – Emphasizing speed and cloud-based innovation
Participated in the CMPE 277 Hackathon, developing a mobile application that integrates cloud services, real-time data processing, and AI-based insights. Designed and implemented scalable architecture, ensuring seamless user experience and efficient backend performance.
COMPUTER VISION RESEARCH
DETR-Demo: End-to-End Object Detection with Transformers
A streamlined object detection framework using Transformers, replacing traditional complex pipelines. Achieves 42 AP on COCO with ResNet-50 backbone, matching Faster R-CNN performance while using half the FLOPs. Implements inference in just 50 lines of PyTorch code.
MULTIMODAL AI
Vision-Language Unified Framework
A novel Vision-Language Pre-training framework utilizing bootstrapping to enhance noisy web data. Achieves SOTA in image-text retrieval, captioning, and VQA, with robust generalization to video-language tasks. Employs a unified encoder-decoder architecture for both understanding and generation tasks.
