Built an intelligent document analysis system with a Gradio UI, featuring a self-correcting RAG pipeline using GPT-4. The tool delivers high-accuracy document Q&A and summarization, ensuring response quality with automated scoring and a guardrail agent

Fine-tuned a large vision-language model (LLaVA-1.5-7B) using Low-Rank Adaptation (LoRA) for medical visual question answering on radiology images. Implemented a custom evaluation pipeline with token-level F1 scoring, managed training workflow on Google Colab with PEFT library, and deployed model artifacts to Hugging Face Hub for reproducible research.

I compared 14 deep learning architectures on the CIFAR-100 dataset, evaluating performance, efficiency, and training characteristics across traditional CNNs, modern architectures, Vision Transformers, and mobile-optimized models.

An open-source package RIDE-CLI is a powerful, user-friendly command-line tool designed to simplify and streamline your data analysis workflow. Whether you're a data scientist, analyst, or researcher, RIDE provides an intuitive interface for exploring, cleaning, and preparing your datasets - all from your terminal!

RIDE is a powerful no-code platform that enables users to upload datasets, preprocess data, visualize it, perform feature selection, explore patterns, and apply statistical tests, all in an intuitive interface of your browser. The platform automatically evaluates model performance, helping users select the best-performing algorithm with ease.

PyTorch Playground, a repository dedicated to showcasing a wide array of PyTorch concepts and operations. It contains hands-on tutorials, code examples, and best practices aimed at helping you master PyTorch from the ground up.

Achieved Top 24% globally in Kaggle competition, implementing Time Series Forecasting models with 98.44% accuracy in predicting microbusiness forecasting.

Deep learning project analyzing La Liga soccer data, featuring interactive visualizations and match outcome predictions using PyTorch. Created a comprehensive dashboard for team performance analysis.

Explored machine learning algorithms for detecting fraudulent transactions using the American Express Kaggle dataset. Implemented XGBoost, CatBoost, and logistic regression for accurate fraud prediction.

Implemented parallelized sorting algorithms and conducted scaling tests on High Dimensional Data, focusing on performance optimization and efficiency.