LDES | Co-Pilot | R&D | Lifecycle Management

Background

Smart and targeted R&D can lead to decreased technology cost and improved quality through product optimization, safety and reliability performance improvements.

R&D can lead to decreased technology cost through the discovery of next-gen materials, design optimization and reliability and safety improvement. Continued R&D funding for mature technologies (e.g., advanced flow battery chemistries) is vital, as R&D advances could contribute 20-35% of the total performance and cost curve improvements. Improving cost efficiency via R&D is especially important for electrochemical and thermal technologies, as many are still in the lab [Liftoff Report].

Challenge

Predicting material properties is a laborious and expensive process, especially when it comes to synthesizing and validating novel materials in the physical world.

Solution

However, there is a promising solution offered by Large Language Models [LLM] & other Neural Networking [NN] techniques, which significantly reduce the time and cost involved in materials development. Through the utilization of LLM & other NN algorithms, inputs such as crystal structure, chemical composition, and electronic band structures can be used to train models that predict computationally challenging or costly properties.

Moreover, Machine Learning [ML] empowers scientists to generate innovative ideas for intricate tasks like solid-state synthesis and crystal structure determination, unveiling previously inaccessible physical insights. This efficient approach allows scientists to explore extensive chemical and configurational search spaces in pursuit of stable, affordable, and high-performance materials.

We are developing an engine, powered by LLM and High Performance Computing [HPC], that facilitates atomistic as well as molecular generative materials design. With the support of Azure computing resources, the engine will be able to seamlessly integrate various stages of the materials design process, including hypothesis generation, calculation setup, experiment execution, informatics analysis, and database development. The laborious and repetitive tasks are significantly reduced, enabling a more efficient and streamlined workflow.

By harnessing the capabilities of supercomputing and leveraging advancements in artificial intelligence, our AI platform will grant access to computed information pertaining to existing and anticipated LDES materials. Additionally, it will offer robust analysis tools that serve as inspiration for the design of innovative materials, ultimately enhancing the quality of your products.

The ultimate goal of the initiative is to drastically reduce the time needed to invent new materials by focusing costly and time-consuming experiments on compounds that show the most promise computationally.

Through the integration of the prime compounds database with generation AI technologies, we can subsequently craft multiple trajectories within the feedstock and design framework. This holistic procedure then transforms into our digital design and intelligent manufacturing methodology.

In conclusion computational materials science is now powerful enough that it can predict many properties of materials before those materials are ever synthesized in the lab.

High Performance Computing

High Performance Computing [HPC] provides the infrastructure that enables your computations, data, and algorithms to run at unparalleled speed in the most efficient manner, saving you time and money.

By scaling LDES materials computations over HPC, we can predict several new battery materials which can enter the lifecycle towards lab testing. There is a wealth of LDES materials out there waiting to be discovered on our way towards liftoff.

Learn more about R&D engine here