Level: Beginner to Intermediate

Prerequisite: None

Norbert Kremer, Ph.D.

CBIP, AICP, DGCP

Cloud Solution Architect

Analytics By Design

Take control of your AI model outcomes by mastering data quality, a critical factor in both training and inference phases. This course offers a deep dive into data quality assessment and improvement practices that drive more reliable, more accurate, and more cost-effective AI solutions.

In this masterclass, Norbert Kremer will examine how data is used across a variety of AI use cases. He will show how AI models with different modalities and scales use training data in very different ways and explain why the large scale of some current models demands new methods of generating training data. Kremer will show that traditional data quality methods developed for tabular data used in BI applications are no longer sufficient. He will show how the economics of building training data sets leads to major innovations in the field.

Using real-world examples, Kremer will demonstrate how data augmentation and generation of synthetic data can improve the performance of AI models by making them more generalizable. He will examine the data-centric movement, including open source and commercial data quality improvement tools. Kremer will show that assessing data quality and improving it as part of iterative AI model development can significantly enhance AI model performance while also managing cost and complexity.

You Will Learn

• How data quality for AI differs from data quality for BI
• Theory of how data is used to train different types of AI models
• Data quality practices and improvement techniques for different types of AI models
• Data-centric AI: stop tinkering with code and focus on data engineering
• How economics drives decisions to collect more data, improve labeling, or improve model code
• The importance of working iteratively to improve data quality, measure performance, and monitor for data drift

Geared To

• Data scientists and AI developers
• Data engineers
• AI engineers
• Business leaders deploying AI models
• FinOps practitioners concerned with AI training costs and AI serving costs and associated unit economics