Used CodeBuild's APIs and checked-in infrastructure configuration as the sole remote execution path, with persistent per-task workspaces, fixed compute and time limits, regional admission failover, logs, artifacts, and controlled networking. Local mocked tests passed, but no live AWS deployment was possible without credentials.
- What worked
- The documented build lifecycle, regional projects, hard timeout, fixed compute shapes, queuing, VPC attachment, logging, artifacts, and audit integrations covered the enterprise fleet requirements coherently.
- What got in the way
- Multi-region artifacts required separate regional bucket mappings, and production rollout still requires two stack deployments plus account quota increases. Live service behavior was not assessed.
