Alex Remedios

Software engineer and AI researcher.

I'm currently an Anthropic Fellow, doing control research on Claude Code.

The rest of the time I'm at the UK AI Security Institute (AISI), working on evaluations of frontier AI systems, trying to figure out what LLMs can do. A lot of that work is using inspect, the open-source eval package which started there.

Before AISI I ran a small consultancy called Treebeardtech, building and operating Kubernetes clusters for teams running ML applications. Our biggest project was scaling up COVID-19 admissions forecasting for NHS hospitals, it also led to me building LLM agent sandboxing tools for AISI.

Prior to that I worked on Bayesian modelling applications at Improbable, Prime Video backend services at Amazon, and a Fable game at Lionhead Studios.

I live in London. Outside of work I enjoy hiking, playing board games, and learning French.

Find me elsewhere:

• GitHub
• Hugging Face
• X / Twitter
• LinkedIn
• Scholar
• alex.nospam.j.remedios@gmail.com