smart_data-1

Smart Data Preparation

We smartly recommend actions to perform on your dataset to amplify hidden signals within your raw data.

smart_data-1
  • AI-POWERED RECOMMENDATIONS

    Get instant, automatic analysis of your data, a rich GUI and one-touch solutions make preparing your data easier than ever.

  • COMPREHENSIVE CATALOG OF ACTIONS

    A catalog of actions supercharges your dataset; from cleaning, structuring, enriching, to engineering new features, so you can seamlessly convert between data types, filter, and reshape your dataset.

  • REUSABLE DATA PREPERATION RECIPES

    Automate your data preparation with powerful, repeatable data recipes that remember actions applied to your dataset.

  • FLEXIBLE FEATURE ENGINEERING

    Quickly generate features required by your models or re-calibrate when you have more training data for optimal predictive insight.

model_recommender-1

Model Recommender &
Performance Prediction

Save time and resources, get recommended the machine-learning algorithm best suited to your data with an automatic view of the models' performance prior to training.

model_recommender-1
  • In-depth performance metrics

    Make the right model selection for your ML objectives, with respect to Training speed, prediction quality, and prediction speed.

  • Predicted performance specific to your data

    Scores are predicted for every model template in our catalog dynamically, as every dataset presents a unique situation.

  • Visual comparison of model templates

    Compare selected model templates easily on a visual GUI, immediately train models with a simple click when satisfied with predicted performance.

model_management

Model life-cycle Management

The one-click deployment automatically turns on monitoring of your model. Data submittedto the model for prediction is automatically logged and checked continuously for drift.

model_management
  • Batch evaluation on new labeled data

    You can perform one-off batch evaluation on new labeled data through the "/evaluation" API call to your model's endpoint

  • Comprehensive Modle Life-cycle Metrics

    Monitor inputs going into your model, as well as prediction outputs generated by your model. Check drift metrics for different types of features individually, or for all inputs.

  • Statistical approach to drift

    Our monitoring functions do not raise a false alarm! With each metric, a significance score is given to help you decide whether an investigation is warranted.