I build governed AI systems that turn strategy into repeatable operating work. My focus is the layer between an impressive model demo and a dependable system: clear ownership, durable memory, deterministic gates, safe tool access, and evidence that the work actually happened.
- RIG Intelligence — a private, modular operating surface for company intelligence, products, departments, and shared agent infrastructure.
- Jake PAI — a local-first operator layer that routes work across people, models, tools, memory, and verification gates.
- Governed agent teams — coordinated Claude, Codex, Hermes, and local-model workflows that share one source of truth instead of producing disconnected copies.
- Evidence-driven execution — ProofPackets, rollback paths, adversarial checks, and explicit approval boundaries for high-impact actions.
Code owns decisions. LLMs assist transformation. Gates decide if it ships.
- Local-first and private by default
- Deterministic before agentic
- One canonical home for every system
- Reversible migrations instead of destructive cleanup
- No source, no number; no proof, no completion claim
- Fewer repositories, deeper modules, clearer ownership
- Consolidating the RIG GitHub estate into fewer than 30 governed repositories
- Preventing agents from writing into duplicate clones or the wrong application
- Building department-level AI employees with scoped tools and durable memory
- Connecting workstation and local-model infrastructure into one auditable fleet
Public repositories are deliberate release surfaces. Internal systems remain private until their security, licensing, documentation, and release boundaries are ready.