Interfaces (ML@CL)
I lead the Interfaces research programme at ML@CL, focusing on interpretable, self-sustaining multi-agent AI systems and extending AutoAI.
The programme takes a systems perspective on the AI adoption problem: software as the interface between socio-technical needs and AI capabilities. It addresses the data dichotomy (data-driven systems must expose data that traditional architectures hide) and intellectual debt in deployed ML systems. Two complementary research lines: DOCS (data-oriented architectures) and S4 (self-sustaining systems with humans in control) support objectives to design, architect, and build AI-based systems and to interpret their autonomous behaviour. Work is validated across different domains, including critical ones like healthcare.
Public software
DOAgent - observable multi-agent systems
iFogSim-Placement - edge service placement simulation (with Joseph Poon, SEAMS 2026 artifact)
Student projects
Part III and Masters project topics at Cambridge are listed in the ML@CL project catalogue.