# Review a GPU cloud for sensitive workloads

A structured review for teams that plan to send prompts, training data or model files to a GPU cloud: data classes, access, tenant boundary, data lifecycle and evidence.

- Level: Specialist
- Length: 60 minutes
- Exam questions: 8, pass mark 80%, up to 3 attempts
- Certificate: valid for 12 months

## What you will learn

- Classify prompts, training data, outputs, logs and checkpoints before upload.
- Ask about scoped credentials, revocation, public addresses and outbound access.
- Ask whether a quoted GPU is dedicated, shared or partitioned, and read Kubernetes GPU settings correctly.
- Check where data lives, how it is deleted and who can reach it, including support staff and subprocessors.
- Match audit evidence to the service and location you order, and record the approval and the exit steps.

## Lessons

1. [Classify data and list artifacts](https://gpuclouds.fyi/academy/review-a-gpu-cloud-for-sensitive-workloads/classify-data-and-list-artifacts.md) (15 minutes)
2. [Access and the tenant boundary](https://gpuclouds.fyi/academy/review-a-gpu-cloud-for-sensitive-workloads/access-and-tenant-boundary.md) (15 minutes)
3. [Data lifecycle and location](https://gpuclouds.fyi/academy/review-a-gpu-cloud-for-sensitive-workloads/data-lifecycle-and-location.md) (15 minutes)
4. [Evidence, approval and exit](https://gpuclouds.fyi/academy/review-a-gpu-cloud-for-sensitive-workloads/evidence-approval-and-exit.md) (15 minutes)

## Questions

### Is this a security qualification?

No. It is a certificate of completion from this site. It does not rate any provider. It does not replace a security review or legal advice.

### Does a vendor audit report settle the review?

No. Check that the report covers the service and the location you order. Then your security and legal reviewers decide whether the evidence meets your duties.

### Do I need to run Kubernetes?

No. One lesson explains how Kubernetes exposes GPUs, because many providers offer it. You can use the lesson to ask better questions even if you do not run it.

Also on the [web page](https://gpuclouds.fyi/academy/review-a-gpu-cloud-for-sensitive-workloads).