Skip to main content
Hollerith makes predictions on tabular data through a scikit-learn style API: fit / predict / evaluate. There is no training step. fit hands the model your labeled table, and predict scores new rows against it in context.

​
Why Hollerith

  • scikit-learn shaped. If you know clf.fit(X, y) and clf.predict(X), you already know the API.
  • No training, no tuning. Point Hollerith at a labeled table and predict immediately. No model selection, no hyperparameter search, no MLOps.
  • Your data is ephemeral. The training table and rows you send are purged as soon as the job finishes; only hashes and counts are retained for metering.

​
What is Hollerith?

Hollerith is a tabular foundation model: a transformer pretrained to predict directly from tables via in-context learning, the same model class as TabPFN and TabICL. Hollerith is trained and served by Monarcha and tuned for larger tables than typical models in this class. See Model for capabilities and known limitations.

​
Next steps

Quickstart

Zero to your first prediction in about five minutes.

Authentication

Create an API key and configure the client.

fit / predict / evaluate

The core workflow in depth.

Model

What Hollerith is good at, and where it is not the right tool.