Successful ML Models.
Every Launch.

SOTAI combines robust interpretable modeling techniques with supporting analysis tooling for results that both you and your boss can understand, control, and trust.

Feature analysis page for a trained calibrated model.

Understand — and control — how your ML models make decisions

Trustworthy Predictions

Incorporate domain knowledge to guarantee real-world relationships between features and targets during the training step with constraints.

Transparency & Performance

Calibrated models are performant without sacrificing transparency. This should be the standard.

Explain Model Decisions

Analyze models at the feature level to explain how your model makes decisions. Even to non-technical stakeholders.

Purpose-Built Tooling

Use purpose-built analysis and collaboration tooling instead of wasting time building custom tooling yourself.

import sotai

df, features, target = sotai.demo.heart()
pipeline = sotai.Pipeline(features, target, "classification")
trained_model = pipeline.train(df)

sotai.api.set_api_key("api_key")
pipeline.analysis(trained_model)
Copy Getting Started Code Button

Getting started is easy

Train a calibrated model on your data in just a few lines of code.

Automatically configure your features and generate SDK code

Automatically configure features with constraints and reasoning.
Then generate the corresponding SDK code and run it.

SDK Code Generation Tool
Generated SDK Code Example

Join our community!

Anyone interested in ML constrained optimization, interpretability, and explainability is welcome!
Let us know how you've been using SOTAI and how it's helped you.

GitHub Icon MarkGitHub Logo TextDiscord Logo