Debapriya Chatterjee

Multimodal LLM-Based Dietary Recommendation System

Graduate Researcher - UConn MINDS Lab

(2026-05)

Multimodal LLM-Based Dietary Recommendation System

• Built Python ETL pipelines to clean, merge and structure 2021-2023 NHANES and USDA FNDDS datasets for downstream modeling
• Designed a multimodal recommendation workflow combining structured health & nutrition data, text inputs, and food-image
• Implemented FAISS-based semantic search to retrieve food-nutrition records for personalized recommendation generation
• Compared LLM-only, RAG, and rule-guided RAG pipelines to evaluate recommendation safety, personalization, and guideline alignment
• Integrated LLaMA Vision outputs with FNDDS food records to support image-based dietary recommendation workflows

Research Assistant - Indian Association for the Cultivation of Science, Kolkata - Kolkata, India

(2022-08 - 2025-08)

• Collected, cleaned, and organized PPMS and MPMS datasets using Excel and Pandas for research analysis
• Analyzed magnetic, thermal, and transport data in Python to identify trends, anomalies and material-property
• Applied linear, nonlinear regression and spline fitting to quantify temperature and field-dependent trends
• Transformed raw instrument output into derived variables for statistical analysis, visualization, and modeling
• Reviewed scientific literature to develop hypotheses, plan experiments and write technical summaries

M.S. - Data Science - University of Connecticut (2026-12)

M.S. - Physics - University of Burdwan (2019)

B.S. - Physics with Mathematics & Chemistry - University of Burdwan (2017)