Leading platform architecture at Sky, currently deep in the agentic-AI track: how do you give internal teams safe, governed AI tooling without it becoming a data-exfiltration story. The syscall-boundary-for-agents idea is getting a real production test.
Standing up the first thin slice of Asklepion Health: tenant isolation primitives, an audited AI gateway, model provisioning, and the regulatory groundwork healthcare demands.
A follow-up to Agentic AI systems on the platforms side. And a longer field report from the edge-GPU years that I've been avoiding because it's the one where I admit what I got wrong.
Going properly deep on eBPF beyond the Cilium surface, writing small programs. The mental model shift from “configure the kernel” to “program the kernel” is the interesting part. Also going deep on medical imaging: computer vision over DICOM, and FHIR for healthcare data interoperability.
Currently in Thinking in Systems (again) and Tidy First?. The full list lives on the bookshelf. Somewhere in the middle of three books, as usual.