Optimize ML costs across cloud providers
skypilot-multi-cloud-orchestrationskillsetup L3★9,423
Orchestra-Research/AI-Research-SKILLs ↗What it does
Orchestrate multi-cloud training jobs across AWS/GCP/Azure with cost optimization
Best for
Running large training jobs by automatically selecting cheapest cloud provider and handling spot preemption.
Inputs
- · training job YAML (code, resource requirements)
- · cloud budget constraints
Outputs
- · job status/logs
- · cost report by cloud provider
Requires
- · sky CLI
- · cloud SDKs (boto3, gcloud)
Preconditions
Cloud credentials configured, SkyPilot installed, job packaged as YAML
Failure modes
- · insufficient quota in chosen region
- · spot instance preemption
- · network latency between clouds
Trust signals
- · multi-cloud cost comparison shown
- · spot instance recovery documented
- · job portability examples