R&D Nerds Run maintains, and the platform work behind it.
A partial list. Entries are labelled by where the work actually happened: R&D we own and maintain, open source we author or contribute to, or employment-era work shipped as a staff engineer inside another company. We don't blur the three.
harness-go
Autonomous multi-agent coding pipeline. NATS JetStream, BadgerDB + Bleve memory, hot-reload policy.
Detailsagentisan
Dual-provider sub-agent primitives for Claude Code and Codex, served over MCP and distributed as an OCI artifact with a 25+ skill library.
Detailstdaddy
Go-native test-impact engine: tells agents exactly which tests a change affects — and which code no test exercises at all.
DetailsOCI skill distribution
Treating AI skills as signed, versioned, governed artifacts delivered to a heterogeneous fleet.
DetailsInternal developer platform
Led at Alaska Airlines: a Backstage-based portal with custom Go services, centralizing onboarding, Kubernetes deployments and service inventory.
DetailseBPF networking migration
Led Cilium adoption and the deprecation of Istio; built zero-downtime cluster-migration tooling across a multi-cloud fleet.
DetailsMulti-region delivery platform
Architected while at GE HealthCare as Senior Staff Architect: a multi-UI platform on AWS with micro-frontends, GraphQL BFFs and multi-region active-active operation. Cross-assistant agentic skills from the same tenure were adopted by 40+ engineers.
Detailsfloci
Upstream work on the 20k-star open-source LocalStack alternative, powering fully local AWS development for enterprise teams.
Detailsstrix_halo_vllm
Ansible collection taking an AMD Strix Halo machine from zero to an operational LLM inference server — vLLM/ROCm and llama.cpp/Vulkan backends.
DetailsFreight Forwarder
Docker image build-and-distribution SDK/CLI from the early container era; used in production at TUNE, published to PyPI.
Details