MANJUSHA PATTADKAL

Proven track record in developing and delivering end-to-end Data Science projects for 5+ years across oil & gas, semiconductors, manufacturing, healthcare and insurance industries. Skilled in Generative AI, Natural Language Processing, Machine Learning and Time-Series Forecasting, with expertise in Python, R, Power BI, and Tableau.

Text-to-SQL: Agentic AI Application

• Developed a multi-agent system using LangGraph and GPT-4o to convert natural language queries into SQL, enabling users to interact with complex databases using plain English
• Designed the system to dynamically handle 15+ database tables, with support for real-time table updates without requiring manual schema adjustments
• Integrated query execution and automated data visualization, providing instant insights through charts and tables based on user input and query results

Recommendation System

• Built an NLP recommendation system to estimate construction project costs by retrieving contextually similar past projects using domain-specific German-language inputs
• Fine-tuned a RoBERTa model for semantic understanding of construction-related texts, and stored dense vector representations in Weaviate for fast and accurate similarity-based retrieval
• Deployed the solution using Azure ML and implemented custom evaluation metrics aligned with human judgment to better assess recommendation quality and contextual accuracy

• Deep Learning Research Assistant
• AI Lab, Otto von Guericke University
• 10/2021 - 01/2022
• Magdeburg, Germany
• Programmed a deep learning model, UNet, using PyTorch for the experimentation of reconstruction of low dose CT-Scan images, with lowering the harmful dosage by 80%, 70% and 55%