Showcasing my best work in AI and ML
Browse the AI/ML projects shown below and explore the complete code portfolio on GitHub.
View all repositories and source code on github.com/ImranRafique847.

Intelligent eBay fashion shopping assistant powered by Corrective RAG (CRAG) on AWS. Users search 18,992 real products using natural language — the AI rewrites queries, performs FAISS semantic search, filters by price/condition, and generates honest responses with product cards.

End-to-end deep learning system classifying car damage severity (minor, moderate, severe, no damage) using EfficientNetB3 transfer learning with 2-phase training. Deployed live on AWS Lambda via Docker container with API Gateway.

AI-powered data analysis agent using Claude via AWS Bedrock that converts natural language queries into automated visualizations. Upload any dataset and ask questions — the agent generates charts and insights automatically.

Conversational AI assistant that controls Gmail, Google Drive, and Google Calendar through natural language. Built with Model Context Protocol (MCP), Claude Haiku via AWS Bedrock, FAISS vector memory, and multi-agent architecture.

Google Cloud Dialogflow CX conversational agent for flight booking. Handles intent recognition, entity extraction (cities, dates, passenger name), confirmation flows, and conditional routing for smooth multi-turn conversations.



Interactive sales analytics dashboard built with Power BI on SuperStore sales data, featuring visual reports, trend analysis, and business intelligence insights.

Machine learning regression model predicting car prices using vehicle features. Includes Jupyter Notebook for EDA and model training, plus a Streamlit web app for real-time predictions.
Salaar Technologies • Onsite • November 2025 – Present
Amazon Web Services • March 2026 – Present
Amal Academy (Stanford-funded) • April 2025 – May 2025
Certura • April 2025 – May 2025