I built an AI-powered real estate research assistant with LangChain, ChromaDB, HuggingFace Embeddings, and Groq LLM. It provides instant, context-grounded answers from multiple URLs and reduces manual research time.
For an e-commerce chatbot for FlipKart, I designed semantic routing between RAG retrieval and natural-language-to-SQL generation. The system supports FAQ and product-discovery queries without manual query writing.
I trained a ResNet50 vehicle-damage classifier using transfer learning, achieving 80% validation accuracy across six categories. I containerized the application with Docker and deployed it to AWS EC2 for real-time inference.
As a Data Science Intern at CodTech IT Solutions, I built and deployed Scikit-learn models for predictive analytics and developed feature-engineering pipelines. I also used Python and SQL for preprocessing and exploratory analysis, and delivered reports and interactive visualizations.
