Inside the AI stack: Vercel agents & GPU optimisation

This is a technical edition of Future Form focused on Vercel’s agent stack. The current features and what’s next on the roadmap. It will also cover how to optimise code to run more efficiently on custom CUDA Kernels.

Luke Philips-Sheard, Engineering Manager, Vercel
Luke will break down GPU performance optimisation and AI sandboxing. Drawing on experience in finance and high-growth environments, he’ll explore how to manage GPU resources, isolate workloads, and prevent performance bottlenecks in real systems.

Jack Foxabbott, Founding Member of Technical Staff, Stealth Startup
Jack will dive into Neural Kernel Search and the evolution of custom CUDA for non-standard architectures. A practical look at how low-level optimisation unlocks performance gains that frameworks alone can’t deliver.

If you’re building AI systems that need to run fast, reliably, and at scale, this one’s for you.

Bonus: pizza, drinks, and practical peer-to-peer AI insights.