Documentation

QED Compute is a managed GPU cloud platform for ML/AI researchers. We handle infrastructure headaches so you can focus on research.

How it works

1. Submit a Research Brief — Describe your project, select GPU type, and upload any code or data.
2. We provision infrastructure — Our team sets up your environment, installs dependencies, and handles configuration.
3. Execution runs — Your job executes on dedicated GPUs. You only pay for actual execution time.
4. Download artifacts — When complete, download your models, checkpoints, and results from the Artifact Vault.

Quick start

1. Sign up and deposit credits
2. Click New Brief in the sidebar
3. Fill out your research requirements and submit
4. Track progress on the My Jobs page

QED uses a prepaid credits system. Deposit funds to your account, and charges are deducted when jobs complete.

Depositing credits

• Minimum deposit: $20
• Preset tiers: $50, $100, $200, $500
• Custom amounts supported
• Payments processed securely via Stripe

What you pay for

Viewing transactions

Visit the Credits page in your dashboard to see your full transaction history, including deposits, charges, and any admin credits.

Choose the GPU that fits your workload and budget.

Multi-GPU configurations

Select multiple GPUs (1-8) when submitting a brief for larger training runs. Billing is calculated as: rate × gpu_count × execution_hours

A Research Brief tells our team what you need. The more detail you provide, the faster we can get your job running.

What to include

• Title — A short name for your project
• Description — What you're trying to accomplish
• Requirements — Dependencies, frameworks, specific versions
• GPU selection — Type and count based on your needs
• Attachments — Code files, datasets, configs (optional)

Tips for faster setup

• Include a requirements.txt or environment.yml
• Specify exact package versions when possible
• Note any special hardware requirements (e.g., NVLink, specific CUDA version)
• Provide a README if your setup is complex

Track your job through each stage of the pipeline.

Billing segments

Each job tracks time in segments. You can view the breakdown on the job detail page:

• Provisioning — Environment setup (free)
• Execution — Active GPU time (billed)
• Intervention — Team fixing issues (free)

When your job completes successfully, artifacts (models, checkpoints, outputs) are uploaded to the Artifact Vault.

Downloading results

1. Go to My Jobs and click on your completed job
2. Scroll to the Artifact Vault section
3. Click the download button to get your files

Balance requirements