Lefts
a domain specific language for building creative ML models
Last checked:
What is Lefts?
Lefts is a domain-specific language (DSL) for composing machine learning workflows that transforms simple models into complex ensembles, cross-validation schemes, and feature engineering pipelines. It works with any scikit-learn-compatible model (those with fit and predict methods) and is built on the Polars DataFrame library.
Lefts pricing
Pricing model: Freemium
Lefts pros
- Composable operations enable building sophisticated ML workflows from simple model components
- Five core operations (Lift, Ensemble, Feed, Tune, Split) cover most common ML patterns
- Works with any model implementing fit/predict interface—no vendor lock-in
- No added complexity to model fitting, evaluation, or experimentation compared to base models
Lefts cons
- Limited to models with fit() and predict() methods; incompatible with some model types
- Requires Polars DataFrame dependency; steeper learning curve for unfamiliar users
- Minimal documentation shown; examples section referenced but not included in this content
Frequently asked questions about Lefts
What types of workflows can I build with Lefts?
Ensembles, cross-validation with different train/test splits, rolling retrains, feature engineering chains (where one model's output feeds into another), and hyperparameter tuning workflows.
What models are compatible with Lefts?
Any model with fit() and predict() methods, such as scikit-learn models (Ridge, RandomForest, etc.), though the underlying model must be Polars-compatible for data handling.
How do I install Lefts?
Install via pip: `pip install lefts`