MaX is on a mission to harness HPC and AI to accelerate the design and discovery of next-generation materials for energy, health, and quantum technologies.
Application sectors: Renewable Energy Materials (lithium-ion and solid-state batteries, high-efficiency photovoltaic solar cells, catalytic materials for fuel cells and electrolysers), Quantum and Nanoelectronics (superconducting qubits, nanoscale quantum sensors, advanced semiconductor materials for next-generation electronics), Biomedical and Molecular Sensing (implantable medical devices, selective gas and chemical sensors, biomolecular sensing materials for diagnostics).
Keywords: AI, Machine Learning, Materials Informatics, HPC, Multiscale Modeling.
The design of advanced functional materials lies at the heart of many of today’s technological and societal challenges. Next-generation batteries, high-efficiency solar cells, biomedical devices, and quantum sensors all rely on materials with enhanced performance and new functionalities. Achieving these properties often requires complex architectures, such as alloys, composites, and heterostructures, that are difficult to model, predict, and synthesize using traditional approaches.
At the same time, the transition toward a sustainable energy economy and the pressure to shorten innovation cycles demand faster and more reliable discovery strategies. For this reason, data-driven methods powered by artificial intelligence (AI) and machine learning (ML) are becoming central tools in modern materials science.
One of the main challenges is navigating the enormous space of possible materials while having only limited experimental data. First-principles methods such as density functional theory (DFT) offer high accuracy, but they are computationally demanding and do not scale efficiently to large or highly complex systems. At the same time, modern high-throughput synthesis and characterization techniques can generate data at a pace that exceeds our capacity to analyze it, creating new bottlenecks in the discovery pipeline.
To overcome these limitations, there is a clear need for integrated AI- and HPC-driven workflows that connect theory, simulation, and experiment. Such approaches can accelerate both fundamental understanding and the practical deployment of innovative advanced materials (IAMs).
The analysis presented in this work shows how explainable AI (XAI) can be used to untangle complex materials datasets. By applying Shapley values, researchers can quantify how individual materials or specific structural features influence model predictions. This makes machine learning models more transparent and reliable, and helps scientists decide which materials should be prioritized for further study or experimentation.
At the same time, machine-learning interatomic potentials (MLIPs) trained on density functional theory (DFT) data enable large-scale molecular dynamics simulations with near first-principles accuracy. These models capture complex behaviors in inorganic, organic, and hybrid systems, while remaining computationally efficient. When combined with adaptive experiment planning in self-driving laboratories, simulations and experiments can inform each other in real time. This approach accelerates the exploration of metastable states, the identification of generative physical models, and the optimization of synthesis routes.
Inverse design strategies further extend these capabilities. Instead of predicting properties from known structures, they work in the opposite direction: they identify candidate structures that meet target properties. This is particularly relevant for optimizing superconducting transition temperatures, designing selective gas-sensing materials, and discovering next-generation materials for energy applications. Overall, these AI-driven approaches lower experimental costs, improve predictive accuracy, and significantly shorten innovation cycles.
The expertise of the MaX team in HPC workflows, materials informatics, and AI-driven modeling plays a central role in these domains. Their experience enables the careful construction and validation of high-quality training datasets, a critical step for ensuring reliable and transferable machine-learning models. The team also optimizes ML models for multiscale simulations, ensuring efficient scalability on modern HPC architectures. By integrating active-learning strategies into the computational workflow, they help create adaptive pipelines in which simulations and data continuously refine each other. This combination of advanced software, high-performance computing infrastructure, and domain-specific scientific knowledge overcomes key limitations of conventional modeling approaches. As a result, predictive materials design could be performed at unprecedented length and time scales, bringing realistic advanced materials within computational reach.
This research presents a unified vision for materials discovery, where AI, machine learning, high-performance computing, and experimental automation work together within adaptive, integrated workflows. By combining ML interatomic potentials (MLIPs), explainable AI, inverse design strategies, and self-driving laboratories, researchers can accelerate both the prediction and the synthesis of innovative advanced materials (IAMs). These approaches deepen our understanding of material behavior across multiple scales and enable targeted optimization for applications in energy, health, and quantum technologies.
Importantly, HPC-enabled AI emerges here not simply as a computational aid, but as a true engine of innovation. It allows scientists to navigate vast materials spaces, account for disorder and complex interfaces, and support near–real-time decision-making in experimental environments. This kind of research improves materials discovery by reducing trial-and-error experimentation and by solving the scalability limits of conventional simulation methods.
The integration of AI-enhanced characterization workflows and interoperable, multiscale digital twins will further shorten development cycles. Such advances will be critical for transferring novel materials from simulation and laboratory validation to industrial deployment.
Talk to us about integrating HPC and AI into your materials research, or consult MaX Software & Data to start accelerating your own discovery pipeline.
Reference paper
Artificial intelligence for advanced functional materials: Exploring current and future directions.