biography
I read Physics at St Catherine's College, Oxford, then spent several years at Philips Research Labs working on gallium arsenide devices, which produced a handful of scientific papers. Somewhere along the way I was a pavement artist in Paris, painting large copies of old masters in acrylic. One of the many highlights was a charming conversation with Bamber Gascoigne in Richmond upon Thames. He even gave me 50p.
An MSc in Computer Graphics at Middlesex took me into visual effects, where I worked as a generalist at MPC and Method Studios on commercials and feature films (IMDb). That led to freelance work in storage and infrastructure, including a spell at DeepMind in its early days, building high-power compute servers for tuning AI models.
Since then I've built cloud web systems using Infrastructure as Code for clients including TUI and the BMJ. I now work in cloud and DevOps engineering, mostly programming in Python.
Throughout this past year I have been developing techniques to use AI efficiently and effectively. I've been through all the pain where the agent must be told things repeatedly and where the context size becomes expensively bloated. I've developed a system to overcome this using strands, so that you don't have to go on the same journey.