Skip to main content
levers9 is the platform behind Levers AI Lab. It runs your Python on a cluster you control: plain jobs, HTTP endpoints, background queues, crons, long-running boxes, sandboxes for untrusted code, JupyterLab notebooks and GPU jobs, all from a decorator and one CLI. On top of the runtime sit two labs. The Data Lab turns recordings and conversations into datasets that people annotate, share under NDA and round-trip through spreadsheets. The RL Arena is where a call-centre team listens to real calls, teaches the agent which reply is better, and ships a new version only when it beats the old one on the same customers.

​
Three layers

Runtime

Jobs, endpoints, queues, crons, boxes, sandboxes, blueprints, volumes, secrets and GPUs. One decorator, the same knobs everywhere.

Dashboard

Apps, containers, workers, storage, secrets, notebooks and model endpoints, live. Workspaces with environments and invited members.

Labs

Datasets with import, pre-labelling and annotation; the RL Arena with call logs, scenarios, side-by-side review, training and evaluation.

​
Start here

Quickstart

Install the CLI, sign in, run your first job, use a GPU.

Every feature

The full map of what the platform does, with a link to the page for each.

Concepts

Jobs, blueprints, volumes, deployments, versions, workspaces, environments.

Demos

One small app per feature, with the exact commands, up to a four-stage L40S pipeline.

CLI reference

Every command, grouped by what you are doing.

RL Arena

Listen, teach, ship: the human loop.