I built and trained a PyTorch CNN freshness classifier, deploying it with Streamlit and improving test accuracy from 88.1% to 94.2% across three model iterations and 28 freshness classes.
At TechAxis, I completed practical training in data analysis, machine learning, and deep learning through projects using Python, SQL, Pandas, NumPy, scikit-learn, TensorFlow, and PyTorch. I also built a FastAPI conversational RAG backend using Qdrant for semantic retrieval, Redis for conversation memory, and Gemini for response generation.
