Data on Demand

Developed Data On Demand (DOD), an AI-powered solution for interacting with Excel, CSV, and SQL databases, using RAG and LangChain memory for personalized insights. Built with Flask, deployed on Azure for scalability, and integrated CI/CD pipelines for continuous updates, improving decision-making.

Pragyan AI Chatbot

Built an Agentic RAG-based GenAI chatbot for complex customer queries, integrating SQLite databases from Firebase and incorporating memory for continuity. Added sentiment analysis to redirect users to human agents. Hosted on Azure, deployed via Twilio API on WhatsApp.

Image Caption Generator

Built a multi-model deep learning pipeline with PyTorch (InceptionV3 for feature extraction, LSTM for captions), integrated DVC and MLflow for MLOps, and deployed using Docker, AWS CodePipeline, and App Runner. Developed CI/CD pipelines and monitored training with TensorBoard, achieving a 0.81 loss

Satellite Image Classifier

Built a multi-class classification model with ResNet18 in PyTorch (96.72% accuracy, 0.09 loss), using DVC and MLflow for version control. Deployed with Docker on AWS CodePipeline, designed CI/CD pipelines, and monitored performance with TensorBoard.

Visual Search & Image Recognition System

Built Visual Search & Image Recognition app with Streamlit. Integrated Annoy Library for faster searches. Optimized preprocessing and deployed on AWS EC2 via Docker. Established CI/CD pipeline using GitHub Actions, with real-time artifact integration from S3.

Cancer Classification Using X-Ray Images

Implemented end-to-end cancer classification project for detecting adenocarcinoma in lung images. Created logging, exception handling, and utility modules. Integrated DVC for version control and MLflow for model monitoring. Deployed Flask-based web app on EC2 using AWS CI/CD pipeline.

Movies Recommendation System

Developed movie recommendation system with content-based and NLP techniques. Modular coding with CI/CD via GitHub Actions and Docker. Deployed on AWS EC2 with Amazon ECR for dynamic updates from S3.

Second Hand Car Price Prediction

Created Flask web app, MLops solution for predicting car prices. Modular coding, MLflow integration, Dockerization. CI/CD pipeline via GitHub to AWS. Hosted on AWS EC2 for global accessibility, showcasing cloud skills.