Mansi A Prabhu

AI and Data Science enthusiast passionate about uncovering patterns in data and turning them into actionable insights. Skilled in building machine learning models and developing intelligent, data-driven solutions to solve real-world problems.

Positive and analytical, with strong foundation in machine learning principles and techniques. Possesses essential knowledge in data preprocessing and model development, with proficiency in Python and TensorFlow. Capable of leveraging technical skills to drive impactful machine learning solutions.

Python, NLP, RAG

• Built a Retrieval-Augmented Generation (RAG) system to summarize and answer questions from lengthy legal documents.
• Implemented document preprocessing, text chunking, keyword extraction and semantic retrieval.
• Generated concise responses from relevant document content to simplify legal information analysis.
• Reduced manual effort required to extract key information from lengthy documents.

Python, Scikit-learn, Pandas, Anomaly Detection

• Developed a machine learning system to identify potentially fraudulent UPI transactions using anomaly detection.
• Performed data preprocessing and analyzed transaction patterns to identify suspicious activity.
• Visualized fraud trends and anomalous transaction patterns to support analysis.
• Detected suspicious patterns across 10,000+ simulated transactions with approximately 86% detection accuracy.

Python, TensorFlow, OpenCV, Swin Transformer

• Developed a deep learning image classification system to identify 5 categories of skin diseases from images.
• Applied image preprocessing and deep learning techniques using the Swin Transformer architecture.
• Built a FastAPI-based prediction API to process image inputs and return disease classifications.
• Achieved approximately 85% classification accuracy on the test data.