Now

I'm focused on a few things right now.

Agentic AI governance and architecture. Most organisations are discovering that their existing AI governance models, built for supervised machine learning, do not hold up when agents start making autonomous decisions. I'm working on the architecture and control patterns that let organisations deploy agentic systems responsibly, at scale, without defaulting to block everything until we understand it. This includes ongoing work with GDS on AI security, and broader advisory on governance frameworks for regulated environments.

AIMux. The open-source agentic AI service mesh I've been developing. The thesis is that multi-agent systems need the same infrastructure primitives that microservices needed: identity, observability, routing, and policy enforcement at the network layer. The current architecture uses a detect-wrap-unwrap pattern with lightweight sidecars for mTLS identity and a gateway that routes on metadata while treating payloads as opaque. I presented the work to the Digital Architecture tribe at Capgemini Invent recently, and I'm building it toward something that can be deployed in real client environments.

Client engagements. Current work includes AI governance and security for UK Government Digital Service, ongoing advisory for A2A Net and Kora Intelligence, and a recent architecture consulting engagement for a former colleague's firm, helping them design a data architecture and AI-powered search platform as they rebuild their technology stack from scratch.

MSc at Bath. The dissertation is forming around bridging the AI strategy-execution gap, the same problem I solve in practice, examined through a more rigorous lens. Modules this year include Advanced Deep Learning and AI as a Social and Political Practice. Expected completion August 2027.

Outside of work, I'm trying to take a proper holiday. Istanbul is high on the list.