My Research

AI systems will soon be generating CAD designs at scale, but currently nobody can verify that a generated part actually does what was asked.^[1]

I'm doing research to make engineered physical systems formally verifiable. The eventual aim is for tools like hardware components or mechanical devices to be derived from formal specifications, just like how a mathematician derives provably correct solutions to equations or makes approximations that are provably close to the true solution.

Who am I?

I'm an Emergent Ventures winner for my research into machine learning in CAD and maths student at the University of Cambridge. I'm a writer on the subject of applied maths and CAD.

My theoretical interests lie within applied category theory, machine learning, topology optimisation and logic. But on this site, I'll often talk about topics completely different to those.

I value the ability to abstract and boil a problem down to only its most essential pieces. I also like to map the structure of analogous domains onto my domain of interest... and you'll see a lot of that in my posts here.

For example, take a look here at the similarities between my work and a 1988 paper on animating Pixar characters.

My current work:

I am building the formal verification layer for CAD.

Contact me:

My email is oli(at)oliverpryce.xyz