✓Transparent onboarding costs and delivery timelines that are met
✓No punitive data egress / offboarding fees
✓Easy onboarding with no preference between console/UI or API/Terraform/CLI, as long as whichever options are offered are well-documented and easy to use
✓Delivery date accuracy and meeting expectations
✓Knowledge of industry experts (e.g., Sylvain)
✓Provisioning of CPU head node for Slurm without explicit requestSlurm
✓Understanding of standard ML user expectations
✓Out-of-the-box GPUDirect RDMA (between NIC and GPU) setup
✓Out-of-the-box IB/RoCEv2 & NVIDIA drivers configuration
✓Performance optimization libraries (e.g., Together AI kernel collection)
✓Audit logs capturing resource actions (create/start/stop/delete), administrative actions (role changes, login success/failure), and billing events
✓Audit log entries include actor identity (user ID, email, IP address), action details, target resource, timestamp, and success/failure status
✓Audit logs queryable via API with filtering by resource type, project, user, and date range
✓Minimum 90-day audit log retention with export capability
✓Audit log access restricted to administrators with no additional usage charges
