Skip to content

Conventions#

Welcome to the LIT platform documentation! To enhance your understanding and streamline your navigation, we’ve organized the content in a slightly unconventional yet intuitive manner. Instead of following the typical left-to-right workflow of starting with data and progressing through to model deployment, we have chosen to approach it from the right to the left. Here’s a brief overview of our approach:

Workflow Overview#

  1. Models First: We begin with an in-depth exploration of models. This section covers model design, compilation, training, evolution, and management within the model registry. By starting with models, we provide a comprehensive understanding of how deep learning models are built, fine-tuned, and evaluated.

  2. Features and Labels: Next, we transition to discussing the features and labels that are fed into the models. We’ll explain why it's crucial to feed features rather than raw data, delve into feature engineering, normalization, and other preprocessing techniques. This section will highlight how these elements contribute to the effectiveness of the models.

  3. Data Management: Finally, we’ll cover data-related aspects—how data is ingested, transformed, and prepared for training. This includes discussions on data acquisition, building training sets, and ensuring that data is effectively structured for model consumption.

  4. Everything Else: Beyond the core components of model design, feature engineering, and data management, the LIT platform offers a suite of additional functionalities to enhance and extend your models. This section covers topics such as deployment, where you’ll learn how to manage and operationalize your models in production environments; insights, including the tools for monitoring and interpreting model performance and predictions; plugins, which allow for custom extensions to tailor the platform to specific needs.

Why This Approach?#

Starting with models helps you grasp the end goal of the AI project—how the data is ultimately used. By understanding the model’s requirements and processes first, you gain clarity on the entire data preparation workflow, making it easier to appreciate the importance of feature engineering and data management.

We hope this approach makes it easier to relate the concepts and appreciate how each part of the workflow connects with and supports the others. As you navigate through the documentation, keep this structure in mind to better understand the interplay between models, features, and data.