Faculty Advisor
Dr. Negin Forouzesh  

Faculty Advisor
Nikhil Dhiman
 

Project Overview

The Amber RAG (Retrieval-Augmented Generation) Project aims to develop an automated knowledge system that transforms decades of Amber Molecular Dynamics mailing-list discussions, tutorials, and documentation into a fully searchable, AI-powered assistant. Combined with a friendly UI new and old researchers will be able to answer any questions they have in minutes rather than hours.

The pipeline ingests raw Amber mailing-list HTML pages, parses them into structured JSON messages, and merges them into topic-level threads. Using embedding models such as MiniLM-L6-v2 and a local database, the system generates vectors that allow users to query the entire archive using natural language. A task scheduler automates scraping, cleanup, validation, embedding generation, and ingestion into the vector database, ensuring consistent updates and long-term maintainability.

The system provides two interfaces: a user interface, where researchers can ask questions and retrieve accurate solutions, and an Aadmin interface, which supports manual queries, data cleaning, and metadata correction. 

Ultimately, the Amber RAG Project enables fast, reliable, high-quality retrieval of molecular dynamics expertise directly from 25+ years of community knowledge without requiring manual searching or external processing.

Team Members

Link: The Amber Molecular Dynamics Package