Cesar Petrescu

Lab ยท AI

  • Internal

Operations automation control plane

A Go orchestrator that separates proposed actions from execution policy

I built a stateless, event-sourced Go control plane for operations automation. An LLM proposes actions, a policy layer decides whether they are allowed, and an append-only audit log preserves the decision history.

  • Go
  • Event sourcing
  • AI agents

Document pipeline

  1. Receive an action proposed by the language model.
  2. Evaluate it against the policy layer before execution.
  3. Record the decision in the append-only audit log.

Outcome

A clear boundary between model suggestions and the authority to act, with an auditable record of operational decisions.