Cairn is a control plane for building, executing, evaluating and continuously improving AI/ML workloads across distributed infrastructure.
Connect data, compute and decisions in a process your team can inspect and improve.
Official focus: ML · Data · Cloud & platform infrastructure
Business rules stay in the domain pack. Vendor connections stay in your deployment settings. Execution support varies by backend.
A fine-tune pipeline as cairn runs it. The first attempt misses the target, so the run is diagnosed and a better candidate goes round again. Only a passing eval reaches a person, and only a person promotes.
Every step is its own governed run: a budget before it starts, a gate wherever a person should decide, a receipt whatever happens. Swap the steps and the same rules apply. The improvement cycle →
Eight ML workflows and nine agentic workflows. Nine industry blueprints to start from.
Train, align, evaluate, compress and run models you own, with the checks built in.
A model investigates, proposes and drafts. A person owns every conclusion.
Reference designs for data-, model- and compute-heavy problems. Start from one; a partner brings the domain model, Cairn brings the governed process.
ML and agentic pipelines marked with a name ship with Cairn. Industry blueprints are reference designs and need a partner with the domain model and data access.
The same fine-tune, seen from the console: the pipeline and where it is waiting, every leg with its cost, the evaluation against its targets, and the decision a person is asked to make.
Seeded demo workspace; run it yourself with cairn dev --seed →
Capture the failure, decide how to respond and evaluate a new candidate. Keep the current approved result until the next one earns its place.
Explore the improvement cycle →Retries address transient failures. Reviewed data and model changes address quality. Each follows its own limits and approval requirements.
No autonomous rewriting of workflows or policy.
On your infrastructure. Docker Compose for a single box, a Helm chart for Kubernetes. SQLite to start, PostgreSQL in production.
Nowhere new. Data stays in your storage (S3, MinIO, GCS, your databases) and jobs run on your engines. Cairn keeps the record: what ran, what it cost, who approved.
Yes. The cairn CLI declares, runs and watches pipelines: cairn pipeline, cairn flow, cairn connect, cairn deploy, cairn worker. Anything the console does, the CLI does.
Apache 2.0 licensed and self-hosted. Early access now; the source opens with the public release.
Put a person in front of the leg that spends money or changes production, and show them the evidence. Let data decide everything a number can decide. How Cairn's gates are built around that rule, and what we deliberately keep away from approvers.
Read the post →How Cairn's fine-tune pipeline prices a run before a GPU starts, checks the trained model against held-out data, and waits for a person before anything is promoted. The commands to run it are at the end.
Read the post →Apache 2.0 licensed, self-hosted. The quickstart gets a control plane with demo data running; a call gets your first pipeline onto it.