Data-Centric AI: Best Practices, Responsible AI, and More

Course details

Machine learning typically focuses on producing effective models for a given dataset. In real-world applications, data is messy and improving models is not the only way to get better performance. Data-centric AI (DCAI) is an emerging science that studies techniques to improve datasets, which is often the best way to improve performance in practical ML applications. While data scientists have long practiced this manually via ad hoc trial/error and intuition, DCAI considers the improvement of data as a systematic engineering discipline. In this course, Aishwarya Srinivasan covers the data-centric principles that guide our path forward in this new age of AI as we shift from a model-centric approach to a data-centric paradigm. Learn about DCAI—what it is and the value it offers. Aishwarya covers the DCAI workflow; MLOps as part of DCAI; data validation and preprocessing; model validation; bias detection and mitigation; responsible AI; and more.

Concepts

0. Introduction

01 - A different approach to AI
02 - Overview of the course objectives and agenda

1. What Is Data-Centric AI

03 - Introduction to data-centric AI
04 - Understanding the role of data in AI and machine learning
05 - Data quality and reliability in AI applications

2. Why Is Data-Centric AI Important

06 - Significance of data-centric AI in real-world scenarios
07 - Benefits of adopting a data-centric approach in AI projects
08 - Case studies highlighting the impact of data-centric AI

3. Workflow of Data-Centric AI

09 - End-to-end workflow of data-centric AI
10 - Deep dive into data-centric AI components
11 - Iterative nature of the workflow for ML applications

4. MLOps - Introduction and Importance

12 - Purpose of MLOps (Machine Learning Operations)
13 - Challenges faced in deploying and maintaining ML models

5. Building Optimized MLOps with Data-Centric AI

14 - Adding data-centric AI principles into the MLOps workflow
15 - Data personas in MLOps workflow
16 - Optimizing the MLOps process - Development
17 - Optimizing the MLOps process - Productionalizing

6. Data-Centric AI in Action

18 - Data validation, train-test validation, and model validation
19 - Best practices
20 - Code example - Exploration

7. Explainability and Interpretability

21 - Importance of model explainability and interpretability
22 - Techniques for understanding and interpreting ML models
23 - Code example - Model validation

8. Bias Detection and Mitigation

24 - Discussion on the challenges of bias in AI systems
25 - Detecting and mitigating bias in data-centric AI projects
26 - Code example - Bias detection and mitigation

9. Data Drift and Model Drift

27 - Monitoring and maintaining ML models in production
28 - Understanding data drift and model drift

10. Responsible AI

29 - Introduction to ethical considerations in AI
30 - Principles for responsible AI development and deployment

Conclusion

31 - Closing remarks and next steps for further learning