Novel AI methods for scientific discovery

Brief description of the context of the research including potential impact: Artificial
Intelligence (AI) offers many opportunities for accelerating scientific discovery in fields like drug
discovery, materials science, climate science, and medical sciences. The use of AI in collaboration
with traditional scientific methods will be crucial as traditional methods often require significant
time and resources, limiting the pace of innovation. In order to solve various scientific challenges
and enable scientific discovery, novel AI methods must be developed. One often cannot simply apply
standard ML techniques to scientific problems, as it requires domain knowledge and understanding
of the data, so innovative methods must be developed. The potential impact of this work includes reducing experimental costs, shortening the discovery lifecycle, and fostering transformative advances
in scientific and industrial applications.
Aims and Objectives:
My primary aim is to develop cutting-edge AI methodologies that drive
forward scientific discovery. One example of this that I am currently working on relates to molecular
design, where my research aims to improve the generation of new molecular structures, accelerating
progress in areas such as drug design and material creation. This research has a few key objectives:
Develop novel AI models capable of proposing innovative molecular structures with particular desired properties
Apply the developed methods to practical use cases in drug discovery and materials science

Novelty of the research methodology: This research will focus on scientific challenges and
the development of novel AI methods to solve them. This novelty will come through exploring advanced AI methods and proposing new alternative methodologies, through exploring fundamental
mathematical and scientific ideas. In the case of molecular design, we are developing an innovative
approach by combining generative models with optimization techniques. This general approach will
contribute to both the theoretical understanding and practical application of AI in scientific fields.

Alignment to EPSRC's strategies and research areas (which EPSRC research area the
project relates to): This research aligns with several of the EPSRC's key strategic priorities,
particularly in the areas of Artificial Intelligence, Computational and Theoretical Chemistry, and
Mathematical Sciences. It supports EPSRC's broader goals of fostering innovation in computational
methodologies and applying them to solve real-world scientific problems. The project contributes
to the advancement of computational methods for chemical discovery, promoting breakthroughs in
Healthcare Technologies and Digital Economy, as well as aligning with the EPSRC's focus on the
use of AI for solving interdisciplinary challenges in science and engineering.
Any companies or collaborators involved: In addition to my two supervisors, Stephen Roberts
and Yee-Whye Teh, current collaborators include Leo Klarner and Tim Rudner