Strategies for Monitoring & Retraining Language Models in Production

In the machine learning model development lifecycle, continuous monitoring and retraining are essential for maintaining optimal performance in real-world applications. In this talk, we review the impact of various decision points on factors like model performance and resource utilization for continuous training of Language Models in production. We have developed a reference framework for designing an effective model retraining strategy to address challenges such as data or concept drift and utilization of the latest data as they become available.

• A framework of decision points to develop a model retraining process.
• Strategies for including newly acquired data for model retraining and evaluation. 
• Empirical evidence on model performance using different retraining strategies.

————————————————————————————————————————————————————

Laura Skylaki – Manager Applied Research | Thomson Reuters Labs

Laura Skylaki is a Manager of Applied Research in Thomson Reuters Labs, where she leads advanced machine learning projects in the domain of Legal and Tax AI. With a career spanning more than a decade at the intersection of research and practical application, she has contributed technical expertise in diverse fields such as bioinformatics and stem cell biology, image processing and natural language processing. She holds a doctorate in stem cell bioinformatics from the University of Edinburgh, UK, and has been publishing on machine learning applications in leading academic journals since 2012.