Welcome to the Adaptive Artificial Intelligence Lab at VUW. We bring together researchers and students interested in machine learning systems that operate over time and in real-world environments. Our work spans areas such as stream and continual learning, multi-agent systems, AI software and open-source frameworks, and applications in domains including energy systems and weather forecasting.
Our Vision
To carry out high-quality fundamental research in machine learning and build open, reusable AI tools that support both research and real-world systems.
About
We study how machine learning systems can learn continuously, respond to change, and remain reliable when deployed in real-world environments. One strength of the group is the development of open-source AI frameworks and research software. These tools support rapid prototyping of new machine learning methods. Related work examines the reliability and security of modern software and AI ecosystems, including software supply chains. We also explore learning and coordination in multi-agent environments, including agent-based models of cooperation. In addition to foundational research, the group works on applications of artificial intelligence in domains such as energy systems and weather forecasting, where learning systems must operate continuously and reliably over time.
Research Themes
- Stream and Continual Learning: machine learning methods that adapt over time to evolving data.
- AI and Software Engineering: reusable open-source tools that support experimentation and rapid prototyping.
- Agentic and Multi-Agent Systems: learning, coordination, and decision-making among autonomous agents, including the use of LLMs for agent reasoning and planning.
- Reliable and Secure AI Ecosystems: research on software supply chains and reproducible systems.
- AI for Real-World Systems: applications in energy systems, weather forecasting and environmental applications.
Why Adaptive AI?
The term adaptive reflects our focus on learning systems that operate over time and must respond to change. Much of our work addresses settings where data distributions evolve, feedback may be delayed or limited, and models must update continuously rather than being retrained from scratch.
In both natural and artificial systems, change is the norm rather than the exception. Intelligent agents intended to operate over long periods must therefore adapt as their environments evolve. 