A Multi-Agent Framework for Assisting Quantum Materials Research

Project phases

Published: December 17, 2025

Last Updated: 1 week ago.

Quantum materials simulation often requires navigating fragmented datasets and manually translating material properties into quantum simulation workflows, a bottleneck for researchers aiming to leverage quantum computing. To streamline this process, the University of British Columbia’s Stewart Blusson Quantum Matter Institute (QMI), in collaboration with the UBC Cloud Innovation Centre (CIC), has developed an AI-powered web prototype that automates the generation of quantum circuits from materials data. The solution enables researchers to move seamlessly from querying materials databases to constructing and simulating Variational Quantum Eigensolver (VQE) circuits and related quantum algorithms.

Building on the foundation laid in Phase 1, the current phase incorporates the AI Strands Agents SDK and Model Context Protocol (MCP) servers to enable dynamic multi-agent workflows. These upgrades support more modular, accurate, and scalable quantum computing pipelines tailored for materials science. In Phase 2, the prototype was developed with an agentic workflow, a multi-agent framework in which a Supervisor Agent orchestrates specialized agents, to support tasks such as multi-material analysis. Providing access to large language models (LLMs), the system integrates real-time data from the Materials Project data and supports quantum computing workflows through both Qiskit and Amazon Braket. The multi-agent orchestration and standardized context exchange architecture reduces time spent on interpreting data for quantum materials discovery, and can also accelerate execution across diverse domains, including enterprise operations, autonomous systems, and data-intensive pipelines.

Approach

The solution is built on serverless architecture, a multi-agent approach designed to simplify how users engage with quantum materials research. Deployed in a Docker-base containerized environment accessed through a Streamlit application, the prototype offers a seamless interface for both users and administrators, connecting securely to the backend via Amazon Cognito.

The system connects to the Materials Project and Amazon Braket MCP servers allowing for dual quantum workflows that bridge quantum algorithm development and hardware execution, enabling researchers to move efficiently from materials analysis to practical quantum experimentation. It provides access to eight AWS Bedrock foundational models to support task-appropriate model selection and cross-model verification, resulting in more precise outputs. Users can conduct operations such as quantum circuit generation, POSCAR file creation, moire bilayer analysis, and integration with simulators and hardware.

Screenshots of UI

This section outlines the core stages of the user journey, as represented through screenshots of the user interface.

USER VIEW – MAIN INTERFACE

Home page after login showing the main interface — _{Upon logging in, users can view the main interface which displays 5 main functions: model selection, framework selection, technical details view, query input and model parameters.}

Features

Model selection dropdown with eight LLM options for quantum and materials science — _{Model Selection: Users can access 8 different LLM models optimized for quantum computing and materials science queries.}

Framework selection dropdown offering Qiskit for VQE circuits and Amazon Braket SDK for advanced algorithms with ASCII visualization — _{Framework Selection: Users can select from 2 quantum workflows: Qiskit for VQE circuits using materials data and Amazon Braket SDK for advanced algorithms with ASCII visualisation.}

Query input field for entering quantum computing or materials science questions. — _{Query Input: Users can input their quantum computing, or materials science, question for analysis by the system.}

Model parameters options for adjusting temperature, maximum tokens, and top-P settings. — _{Model Parameters: Users can adjust the behaviour of the model by changing parameters such as temperature, maximum token count, and top P which controls the diversity of word selection.}

Technical details view showing MCP tools, Braket processes, agent workflows, and API calls — _{Technical Details View: Users can view background processes such as MCP tools used, Braket processes, agent workflows and API calls.}

ADMIN VIEW

Admin panel for creating user accounts and assigning admin privileges — _{Administrators can create new user accounts, as well as assign admin privileges to users.}

Admin panel showing a list of users with roles, creation dates, and options to delete accounts — _{Administrators can view all users, their roles, and creation dates. They can also delete users.}

Admin panel for assigning and managing admin roles and viewing users with admin privileges — _{Administrators can assign and manage admin roles, and view all users with administrative privileges.}

Supporting Artifacts

Click below to see technical details of the solution, including the detailed Architecture. Or click here to go directly to the project GitHub repository.

Architecture Diagram

A detailed explanation of the diagram can be found on the project GitHub repository.

Technical Details

The Quantum Materials Code Generation prototype enables quantum computing and materials science research through intelligent Large Language Model interactions on secure AWS infrastructure. It is built with scalable cloud architecture and dual quantum computing frameworks for comprehensive research workflows.

Users can interact with quantum algorithms and materials data through a Streamlit web interface that automatically processes queries using AI Strands Agents SDK multi-agent architecture. The Supervisor Agent orchestrates specialized agents (DFT Agent, Structure Agent, Agentic Loop) that connect to Model Context Protocol (MCP) servers for enhanced data integration. It provides access to eight AWS Bedrock foundational models: Claude Sonnet 4.5, Claude Opus 4.1, OpenAI OSS-120B, Qwen 3-32B, DeepSeek R1, Nova Pro, Llama 4 Scout, and Llama 3 70B. These agents leverage Materials Project MCP server for crystallographic data and Amazon Braket MCP server for quantum device information, enabling dual quantum workflows: Qiskit framework integration with Materials Project data for VQE circuits and Amazon Braket SDK for advanced quantum algorithms with ASCII visualization.

The solution deploys on AWS Elastic Beanstalk with Docker containerization. CloudFront delivers content with SSL termination for secure access. Amazon Cognito provides enterprise authentication with admin-controlled user creation. AWS Secrets Manager securely stores API keys. It integrates with quantum simulators through Amazon Braket Service. This supports operations ranging from creating quantum algorithm examples to production quantum computing workflows.

Acknowledgements

This project was created in collaboration with the Stewart Blusson Quantum Matter Institute (SBQMI).

Image from the SBQMI.

Student Team: Development by Sharon Marfatia. Project assistance by Anya Ameen.

About the University of British Columbia Cloud Innovation Centre (UBC CIC)

The UBC CIC is a public-private collaboration between UBC and Amazon Web Services (AWS). A CIC identifies digital transformation challenges, the problems or opportunities that matter to the community, and provides subject matter expertise and CIC leadership.

Using Amazon’s innovation methodology, dedicated UBC and AWS CIC staff work with students, staff and faculty, as well as community, government or not-for-profit organizations to define challenges, to engage with subject matter experts, to identify a solution, and to build a Proof of Concept (PoC). Through co-op and work-integrated learning, students also have an opportunity to learn new skills which they will later be able to apply in the workforce.

Projects