My Projects
Welcome to my projects page. Here, you can explore the work I've done in the past. Each project is a testament to my dedication and expertise in the field. Feel free to take a look around and get to know me better.
Integrated Data Analytics and Time Series Forecasting for Aquaculture
Time series forecasting, Predictive Modeling
Developed an AI-driven aquaculture system at HBOI, integrating IoT sensors and machine learning to boost productivity by 3% and reduce waste by 15%. Achieved 91% accuracy in demand forecasting, analyzing data from 10+ species to optimize sustainable fish farming operations.
AI Assistant for Personal Data
NLP, LLM, Retrieval Augmented Generation
This project offers an AI assistant for analyzing personal data files, implemented in two ways: a Jupyter notebook using Hugging Face models and Gradio, and a Python script using IBM Watson ML and Flask. Both versions allow users to upload documents, ask questions, and receive AI-generated answers based on the content.
Graduate Teaching Assistant Management System
Python Full-stack Webapp with NLP features
GTAMS is a Flask-based web application for managing Graduate Teaching Assistants in universities. It features user authentication, application submission, NLP-based resume parsing, and TA evaluation tools, streamlining the GTA management process for administrators, faculty, and students.
Non-Toxic Summarizer
NLP, LLMs, Fine-Tuning, PEFT
This project fine-tunes a FLAN-T5 model using Reinforcement Learning to generate less toxic summaries. It employs a hate speech reward model and Proximal Policy Optimization (PPO) to reduce output toxicity while maintaining summary quality.
French Buddy
NLP, LLM inference webapp, Prompt Engineering
This project is a web-based English-to-French translator using IBM Watson's Machine Learning and speech services. It features speech-to-text input, LLM-powered translation, and text-to-speech output, built with Flask to assist French language learners.
AI Personal Voice Assistant
NLP, LLM, OpenAI GPT 3.5, IBM STT & TTS
This project is a voice-activated AI assistant using OpenAI's GPT-3.5 model. It features speech-to-text input, natural language processing, and text-to-speech output, creating a seamless voice interaction experience. Built with Python, Flask, and web technologies, it offers a responsive interface with customizable voice options.
Resume Keyword Assistant
LLM inference webapp, Prompt Engineering
Resume-Keyword-Assistant uses the Gemma 2B language model to help job seekers tailor their resumes. It analyzes uploaded PDF resumes and suggests ways to naturally incorporate specific keywords, improving relevance for targeted job applications. The tool features a user-friendly Gradio interface for easy interaction.
Meeting Summarizer
LLM, Gradio, Speech to Text Summarization
Meeting Summarizer and Speech Analyzer, transcribes audio using OpenAI's Whisper and summarizes content with a BART-based model. It features a Gradio interface for easy use, supports GPU acceleration, and offers both Jupyter notebook and Python script implementations, with the latter using IBM Watson ML for language processing.
Automated Image Captioning
BLIP Transformers, Gradio
Automated Image Captioning using Transformers, offers multiple implementations for generating image descriptions. It includes scripts for URL-based and local image captioning, a Gradio interface, a BLIP model implementation, and a Flask web application. The project leverages various deep learning techniques to provide comprehensive image captioning solutions.
Song Lyrics Generator
LLM, Fine Tuning, PEFT, Web Scraping
LLM-Punjabi-song-lyrics-generator uses GenAI to create Punjabi lyrics in the style of a late artist. It employs web scraping for data collection, PEFT techniques for model fine-tuning, and a GPT-2 base model. The system generates original lyrics using a text generation pipeline implemented in Python with PyTorch and Transformers libraries.
Data Analysis Webapp
LLM, Fine Tuning, PEFT, Web Scraping
Data Analysis Webapp is a Flask-based platform for IoT systems managing biomass sensors and HAUCS. It features Python scripts for routing and data visualization, integrates with Firebase for data storage, and includes both real sensor data and generated datasets. The project uses HTML, CSS, and JavaScript for a responsive web interface displaying graphs and analytics.
Falcon9 Launch Capstone
LLM, Fine Tuning, PEFT, Web Scraping
This project, Falcon9 Launch Data Science Capstone, aims to predict the successful landing of SpaceX Falcon 9's first stage. It analyzes a dataset of over 100 launches from 2010-2022, including launch details and outcomes. The goal is to develop a predictive model to estimate launch costs, potentially aiding competitors in bidding against SpaceX for rocket launches.
An Artificial Intelligence Driven Framework for Classification of Ophthalmic Images Using Convolutional Neural Network
Computer Vision, Deep Learning, Classification Models, Neural Networks, CNNs
Led a CNN-based project achieving 98% precision in diagnosing multiple eye diseases. Optimized advanced neural network architectures, improving clinical efficiency by 40%. This work demonstrates my expertise in applying AI to solve critical healthcare challenges and enhance patient care.