AION Engines

• INGESTOR (Data Ingestion): Data is ingested into AION from various sources.
• EXPLORER (Exploratory Data Analysis): The nature of data, feature relationships, model statistics, and performance information are shown to provide insights.
• TRANSFORMER (Data Processing): Data cleaning, preparation, and outlier detection are automated to enhance data quality and improve model accuracy.
• SELECTOR (Feature Selection): Statistical analysis identifies and keeps relevant features for model training, while removing unimportant features based on correlation and importance.
• LEARNER (Model Training/Hyperparameter Tuning): Configured models are trained, and the best parameters are selected based on hyperparameter tuning. Many algorithms are supported.
• PREDICTOR (Inference Service): Handles ML model serving and inference.
• OBSERVER (Model Monitoring): Monitors the model for input and output drift.
• EXPLAINER (Explainable AI): Explains the model and prediction uncertainty.
• CONVERTOR (Model Conversion): Converts generated models into formats suitable for embedded, cloud, or edge deployments.
• TESTER (Model Testing): Employs different testing methodologies to test the generated models.

• CODER (Machine Learning as Code): Automatically generates Python code for the ML pipeline components.