Overview
The discovery of novel inorganic solid-state materials is essential to advance energy storage, catalysis, semiconductors, and quantum technologies. However, the design and discovery of new materials remain major scientific challenges, and this project aims to address these challenges by developing generative machine learning and AI models to improve crystal structure prediction workflows for inorganic solids.
About this opportunity
The project will explore cutting-edge techniques in generative AI modelling (e.g., reinforcement learning, diffusion models, LLMs) to predict new structures. These models will be integrated with chemically informed constraints and first‑principles calculations to generate novel crystal structures. The generated structures will be validated using physics‑based simulations and benchmarked against major materials databases to assess accuracy and novelty. The overarching goal is to create a computational workflow capable of proposing structurally novel experimental targets to enable a step change to AI-accelerated materials discovery.
The research direction of this ambitious project can be shaped by the student’s own scientific intuition and creativity, evolving to meet the project goals and with direction from a multidisciplinary research team.
You will have access to high-performance computing resources, work closely with experimentalists, and have the opportunity to publish in leading journals. This studentship is suited for a student with a background in computational materials science, machine learning or artificial intelligence. Experience with Python and writing code is essential. Experience with ML frameworks (PyTorch/TensorFlow), graph and/or neural nets and familiarity with materials science, crystallography and/or solid-state chemistry would be an asset. Please clearly highlight your relevant experience in your application.
Further reading
1. Discovery of Crystalline Inorganic Solids in the Digital Age. D Antypov, A Vasylenko, CM Collins, LM Daniels, GR Darling, MS Dyer, JB Claridge, MJ Rosseinsky, Acc. Chem. Res. (2025), 58 (9). pp. 1355-1365. 10.1021/acs.accounts.4c00694
2. Integration of generative machine learning with the heuristic crystal structure prediction code FUSE, CM Collins, HM Sayeed, GR Darling, JB Claridge, TD Sparks, MJ Rosseinsky, Faraday Discuss., (2024), 256. pp. 85-103. 10.1039/D4FD00094C.
3. Superionic lithium transport via multiple coordination environments defined by two-anion packing, G Han, A Vasylenko, LM Daniels, CM Collins, L Corti, R Chen, H Niu, Hongjun, TD Manning, D Antypov, MS Dyer, J Lim, M Zanella, M Sonni, M Bahri, H Jo, Y Dang, CM Robertson, F Blanc, LJ Hardwick, ND Browning, JB Claridge, MJ Rosseinsky, Science, (2024), 383 (6684). pp. 739-745. 10.1126/science.adh5115
4. Introducing physics-informed generative models for targeting structural novelty in the exploration of chemical space, A Vasylenko, F Ottomano, CM Collins, R Savani, MS Dyer, MJ Rosseinsky, (2025), 10.48550/arXiv.2510.23181