At Accenture, I built an AI-powered image metadata platform using FastAPI and Streamlit, integrating BLIP, Qwen2-VL, and Azure AI Vision for captioning, semantic tagging, model evaluation, and batch processing.
I also developed OCR and product information extraction workflows with Azure AI Vision, EasyOCR, OpenCV, and ZXing. Using an Ollama-hosted Qwen2.5-3B model, I inferred target audience segments from image metadata and produced structured CSV and JSON outputs.
For my Amazon Clone project, I built a full-stack e-commerce application with user registration, product catalog, shopping cart, order processing, and RESTful APIs. Its React frontend supports multiple pages, while JWT authentication handles user sessions.
I also built a Java autocomplete engine using a Trie, priority queues, and frequency tracking, and developed AutoVal, a Flask app that predicts used-car prices with a Scikit-learn and Pandas pipeline.
