Skip to main content
swale

Resource-heavy workflows,
zero-effort infrastructure.

Swale runs the workflows that need serious resources — AI model pipelines, video processing, large-scale research. Define the compute each task needs, get dedicated storage at every level, and let the platform handle provisioning, scaling, and cleanup.

Platform

Big resources, only where they're needed.

A Swale workflow is a set of tasks. Each task declares the compute it needs, so only the work that truly needs multi-GPU machines runs on them — everything else stays lean. Storage works the same way: a massive persistent store for your account, plus dedicated ephemeral space for every workflow and every task.

Per-task compute

Declare the compute each task needs — CPU, memory, or multi-GPU machines. Only the tasks that genuinely need heavy hardware get it, so you never pay to keep a GPU idle.

Massive per-account storage

Every account gets an extremely large persistent store that lives across runs — a permanent home for your datasets, models, and results.

Ephemeral storage per workflow

Each workflow run gets its own dedicated, large ephemeral volume — shared scratch space for the tasks that make up the run, cleared when it finishes.

Ephemeral storage per task

On top of that, every task gets its own dedicated large ephemeral volume for intermediate work, cleaned up automatically when the task ends.

Logs and usage, built in

Every run captures task logs and overall resource usage out of the box. See what each task did and what your workflows consumed — no extra tooling to wire up.

Zero-effort compute and storage

No clusters to size, no volumes to provision. Describe the work and Swale allocates the compute and storage each task needs, then tears it all down after.

Zero effort, for both compute and storage.

No clusters to size. No volumes to provision. Describe your workflow and its tasks, and Swale allocates exactly the compute and storage each one needs — then cleans everything up when the run finishes.