AI Systems Lab
Intensive applied programme
People designing AI products, copilots, workflows, and agentic systems.
Build more useful, secure, auditable, and human-governed AI systems.
Authority & provenance
Who may act. What an input becomes after it is transformed. How to keep data from becoming a command.
Tool design
Least privilege, typed intents, spending and blast-radius limits, structured protocols.
Human checkpoints
When fluency is not enough. Designing review that people will actually use.
Audit & red team
Authorised testing of your own agent. Residual risk you can show an employer or a regulator.