I've designed, built & scaled a bunch of distributed storage systems at several growth-phase startups over the years; see my personal site for more info.
Currently focused on Invar, a diskless transactional document store. Invar is complemented by managed cloud and enterprise offerings. Please feel free to give it a star or join our Discord!
- Golang / Rust / Java (compiler internals; bytecode manipulation) / JVM-hosted languages, esp Clojure and JRuby
- Distributed systems
- Networking: async/event-driven I/O; L4/7 protocols (especially HTTP and QUIC)
- Linux I/O internals
...plus the usual table-stakes stuff: shell scripting, containers & container orchestration (k8s), CI/CD
My main contact points with AI come in 2 forms:
- Agentic development: My stack of LLMs and associated tooling is quite modest; for me, systems programming is quite frugal token-wise but doesn't easily lend itself to fully closed-loop development or self driving codebases.
- Vector storage, with a focus on best-practice implementations over LSM-Trees (primarily for the benefit of Invar)
- Invar: Diskless transactional document store with commercial support.
- Swift-Disruptor: Swift port of LMAX's concurrency library. Used internally at Apple.
- Cups: Formerly the canonical way of calling printers from Ruby; I no longer have time to maintain it and several forks have spun off. So I'll claim the legacy, even if I'm not necessarily proud of the code.