Open this Iris pipeline
Sample already loaded. Click Run on the canvas. Upload of your own file stays locked until you sign up.
Loading…
Visual sklearn for CSVs — copilot builds the DAG, you keep the code.
50 jobs / month, 100k rows, no credit card.
Sample already loaded. Click Run on the canvas. Upload of your own file stays locked until you sign up.
Three ways in
Train Random Forest on Iris, see hold-out metrics, then export labels or probabilities.
Try sample
Run K-Means on a sample table. Inspect the plot, then swap in HDBSCAN or GMM.
Run K-Means on Iris
Copilot draws the DAG. You keep the nodes, the metrics, and the Python.
Open canvas + copilot
White-box on purpose
Researchers bounce from black-box SaaS. Hold-out is labeled separately from training scores. You can leave with Python.
Optimistic training scores are marked. The number you ship is hold-out.
Export P(class), not only Yes/No. Download the ROC/PR curve.
from sklearn.cluster import KMeans pipe.fit(X) labels = pipe.predict(X)
Export a sklearn script or notebook. The DAG is the artifact, not a chat log.
Your file is used to run your job. Sign up when you want to upload it, save the graph, or download exports.
ModulesDedicated pages for each family — also under Modules in the top bar, including from the canvas.
Group similar data points together without labels. Discover hidden patterns, segment customers, detect anomalies, and explore the natural structure in your data.
Unsupervised Learning · 14 algorithms · Open clustering
Train models to predict categorical labels from features. Build classifiers with confusion matrices, ROC curves, feature importance, and calibration diagnostics.
Supervised Learning · 9 algorithms · Open classification
Predict continuous numeric values from features. From linear baselines to gradient boosting — with R², residual plots, and feature importance diagnostics.
Continuous Prediction · 11 algorithms · Open regression
Clean, transform, and encode datasets through a sequential Transformation Chain. Impute missing values, scale features, encode categoricals, and filter columns — then export to Pipeline or Python.
Data Preparation · 20 methods · Open preprocessing
Reduce high-dimensional data for visualization, noise removal, and feature extraction. From classic PCA to modern manifold techniques like UMAP and t-SNE.
Transformation & DR · 8 algorithms · Open dim. reduction
Chain preprocessing, models, and plots. Ask copilot in English or drag nodes yourself.
Canvas builder · visual + copilot · Open canvas
Free account. No card. Export labels, probabilities, and ROC. 50 jobs / 100k rows a month.
Signup unlocks upload, save, and download — samples stay free to run