Discovery

10,000-foot view

The LIT feature store is engineered to handle models built on exceptionally large datasets, providing powerful tools for efficient data exploration. Users can interact with datasets containing billions of data points with ease, thanks to rapid random access capabilities that deliver results in under 30 milliseconds. The reference architecture leverages WebAssembly technology to power a responsive chart view, which allows users to zoom in and out of data seamlessly.

This combination of fast data access and dynamic visualization enables users to navigate large datasets efficiently and analyze extensive data in real time. With these capabilities, LIT ensures that users can swiftly explore and interpret vast amounts of data, facilitating more effective model training and insights.

Visual

The visual representation of features aids in feature development by making it easier to identify and interpret data patterns and trends.

Outliers

Utilizing the correct features is critical to solving machine learning problems. When designing a new feature, Lit allows you to instantly preview it across the entire dataset, allowing rapid iteration of that feature in order to solve problems before costly data builds or wasted time training fundamentally flawed models.