Idiology

This section will describe the core ideas of Aligned, and explain why it differs to other approaches.

Describe What not How

One of the core ideas is that by describing what we have to work with will rather then how.

This may seem a bit odd at first, as we do not describe any clear use-cases. We only describe what our features are, and where they are located. Rather, our use-cases is done later in application spesific code, and then we provide the definitions that we will work around.

Therefore, we can make our code more flexible, and adapt our businiss logic to a wider range of scenarios.

Here are a list of use-cases that we can easily fulfill.

• Reuse transformations for different types of data sources and processing engines - sql, pandas, polars, spark, etc.
• Load feature for batch prediction
• Load features and ground truths for training
• Process features using stream processing for predictions
• Setup low latency data sources with cached data
• Setup data cataloges for documentation and discoveries
• Load predictions and ground truth for model performance monitoring