Infrastructure & agent systems
Infrastructure & agent systems for teams that outgrew the defaults.
2 live products on a self-hosted NixOS/Kubernetes
fleet.
Previously reduced a 66,000-line platform to its
durable core and shipped semantic retrieval the next day.
I build self-hosted platforms and AI agent infrastructure — and the APIs that agents consume. Then I make them boring to operate.
max@ronner.devSelected work
-
JobFilter.ai active
Pivoted from a full career platform to a focused API; semantic retrieval shipped the next day.
Public-job corpus API for agents: structured JSON only, source attribution, tokens scoped to
jobs:read. In July I deleted 66,000 lines and kept the layer that proved durable. -
Agent permission model published
Replaced runtime approval dialogs with a build-time proof.
Permissions for AI agents as a build-time proof instead of a runtime habit: capabilities → scopes → profiles in home-manager.
nixos-rebuildfails naming unknown capabilities before anything runs — the sandbox profile gets a read-only project and four tools. -
Cluster & build farm running
Self-hosted deploy pipeline that ships this site from actual hardware.
2-node Kubernetes on mini-PCs, a Nix build farm across it, Flux image automation, everything versioned. Parallel rolls once hit memory pressure mid-deploy; rolls are sequential now — slower, deliberately.
If you're building agent infrastructure or self-hosting critical systems, describe the failure mode, not the stack.
max@ronner.dev