Taking Machine Learning Research to Production - Solving Real Problems
Explore the challenges and solutions for transitioning machine learning research into production applications in this GOTO Copenhagen 2019 conference talk. Learn about ML pipeline architectures, particularly Google's TensorFlow Extended (TFX), for implementing scalable and robust ML applications. Discover how to address issues unique to ML and data science, including testability, scalability, training/serving skew, and component modularity. Gain insights into measuring model fairness and predictive performance across user segments. Understand the importance of software development methodology in ML applications and how TFX enables strong practices such as testability, hot versioning, and deep performance analysis. Benefit from Google's experience in using TFX for large-scale ML applications and learn how developers can effectively move their ML projects from research to production environments.