Christian Browning

Project: Guaranteed structure prediction with machine learning

Supervisors: Vladimir Gusev, Matt Rosseinsky

What inspired you to pursue this project and join the DAMC CDT?

I studied Computer Science at the University of Liverpool, where I developed a strong interest into artificial intelligence, optimisation, and mathematically rigorous approaches to difficult problems. What made me want to specifically do a PhD was the opportunity to work on research that is technically challenging and also capable of contributing to real human progress. I was especially attracted to this project because crystal structure prediction is already a well-established and highly populated research area, but this project approaches it in a very unusual way. I was drawn to the use of integer programming and rigorous optimisation methods, particularly because they can provide guarantees and certificates of optimality in a way that most machine learning systems cannot. I also liked that the work focuses on developing better algorithms, rather than automatically defaulting to the current deep learning-heavy research culture. I think there is real value in exploring strange, contrarian, or less fashionable approaches when they open up new parts of the research space and lead to genuinely different insights. The DAMC CDT stood out to me because of its interdisciplinary nature. It offers the rare opportunity to work across computer science, materials chemistry, and automation, while being part of a cohort focused on the full pipeline of discovery, from theory and prediction through to synthesis and process automation.

What is your research project about, and what impact do you hope it will have?

My research focuses on crystal structure prediction using discrete optimisation methods, with the aim of making computational materials discovery more rigorous, reliable, and explainable. In particular, I am interested in combining optimisation, structured search, and constraint-based reasoning to improve how candidate crystal structures are generated and validated. A distinctive part of what I bring to the project is my background in LLM orchestration and AI systems design, which opens up opportunities to build more intelligent research workflows around optimisation-based materials discovery. I hope this work will contribute to methods that not only predict structures efficiently, but do so with stronger guarantees and clearer reasoning than conventional heuristic approaches alone. In the long term, I hope it helps accelerate the discovery of useful new materials and strengthens the connection between computer science and chemistry in scientific research.

What has been the most exciting or rewarding part of your PhD journey so far and how does your project benefit from being part of an interdisciplinary CDT? 

One of the most rewarding parts of the PhD so far has been being part of an interdisciplinary CDT, where I can see the wider research space from multiple angles. It has been especially valuable to learn from people working at different stages of the pipeline, from computational discovery and prediction through to synthesizability, experimental constraints, to finally process automation and optimization. That perspective helps ground the computational side of my own work in the real costs, challenges, and decisions that come after prediction. Being part of DAMC makes the project far more ambitious and meaningful than it would be within a single department as I am closer to the real cost and need for my work.