Projects
A collection of projects showcasing my work in AI/machine learning, cloud architecture, and web development from my research, coursework and personal collection.
Personalized Writing Assistant
Fine-tuned Open-source Large Language Model (LLM) to adapt to personalized writing styles and preferences. Utilizes Llama and Mistral models, fine-tuned on custom datasets of writing samples from various domains such as emails, blog posts and reviews.
Documenting Human Behavior in Videos
Developed a tool to analyse human behavior in videos for clinical diagnostic applications using GPT-4o for multimodal video and language understanding.
Time Series Forecasting of Store Sales
Building a forecasting model that addresses seasonality and trend while predicting store sales using time series features such as sales data, holidays, and transactions. The analysis involves hypothesis testing, data preprocessing, feature engineering, and applying both machine learning & statistical models such as ARIMA, Exponential Smoothing, XGBoost, LightGBM, and Prophet for prediction.
Stats-aware GANs for Image Synthesis
Generating statistically accurate synthetic images for domain-specific applications like medicine and material science. We extended StyleGAN2 by integrating new regularization loss functions—based on graph-based statistical tests and distributional similarity metricsto improve the statistical alignment between real and generated images, ensuring more reliable synthetic data for specialized fields.
VisionAid: Assisted Living Tool for the Blind
A computer vision based smart home solution for finding household items that uses a YOLO-v4 object detection model trained on a custom dataset of common household items, intergrated with a voice-enabled web application for accessibility.