Our vision is to advance fundamental underpinning AI methods and build solutions for critical societal problems. This includes health, discovery of new materials for greener technologies, aged care and security.

120+

AI researchers & software developers

600+

top-tier publications

50+

collaborative partners

6

PLOPS compute infrastructure

‘We collaborate with many disciplines, and solve problems at the nexus of computer science and the other disciplines. This has allowed us to do really innovative work. I’m deeply grateful to a wonderful team of people who I have worked with over the years who have made this possible.’

AI Advances for Industry

Health

Sustainability

Defence

Manufacturing

Education

Space

Research themes

Deep learning, computer vision

We advance neural architectures for perception, reasoning, and interactive collaboration. Our work includes recurrent memory systems, visual cognition and other innovative concepts that lead to powerful intelligent systems tackling problems in industry and society.

Bayesian optimisation

Optimising neural networks to make them more efficient — reducing expensive network costs by increasing performance, which means your networks require less resources to achieve more advanced results.

AI safety and assurance

Identifying efficient ways for machine learning to occur with minimal human supervision — solving critical real-world problems such as drug design, chemical reactions, and more.

Reinforcement learning

We design adaptive systems that learn from interaction, including simulations to real-world (sim-to-real) transfer, robot design, and vision-language-action pipelines. Our focus is on creating systems that can adapt efficiently and reliably in changing contexts.

Agentic AI

We address crucial issues to create capabilities in agentic AI systems, and this includes theory of mind, empathy and situational awareness. We create agents that can work with multiple modalities like vision, text and speech and focus on aspects that build effective AI agent and AI/human teams.

Generative AI and foundation models

We work on the development and analysis of language and generative models, from diffusion and small language models to foundation-scale systems. Our interests include model safety, alignment, and interpretability, as well as cognitive architectures that incorporate memory and reasoning for more reliable outputs.

Human-machine teaming

We develop AI systems that can work in close partnership with their human users enabling shared autonomy enabling growing understanding of the partnership and a deep understanding of each partner’s expertise. Estimating and communicating intent is a core aspect.

Probabilistic methods

We develop cutting-edge probabilistic models to estimate model uncertainty to ground and support principled decision-making in complex environments.

Sample-efficient optimisation

Our team has deep expertise in Bayesian optimisation, A/B testing, and multi-armed bandits, aiming to accelerate the experimental design process when data is limited, thus being able to optimise both the constituents of a product and the process by which it is made. Novel material discovery is an important application area.