← All programs
Applied AIAdvancedACD-02
Hosting and supervising AI agents
The infrastructure half of the AI conversation: where an agent runs, what it is allowed to touch, how you watch it, and how you stop it when it drifts.
What you leave with
- Design an isolated execution environment for agents
- Scope tool access and secrets per agent identity
- Instrument behavior, spend and drift
- Write a rollback and kill-switch procedure that works
Modules
- 01
Anatomy of an agent
Loop, memory, tools, and where things go wrong.
- 02
Isolation and sandboxing
Containers, gVisor, network egress control.
- 03
Identity and tool scoping
Least privilege for non-human actors.
- 04
Observability
Traces, decision logs, cost and latency budgets.
- 05
Guardrails and human gates
Approval checkpoints and refusal handling.
- 06
Failure day
Lab: contain an agent that has done something wrong.
Who it is for
Cloud and platform engineers already running agents or about to