Joel Karpiak

Company: GlaxoSmithKline

Job title: Head of Protein Design and Informatics

Bio:

Joel joined GSK in 2015 from UCSF, where he focused on large-scale modeling and virtual screening for orphan GPCRs. Initially working in small molecule design for over 5 years, he transitioned over into growing the Protein Design & Informatics group for the past 3 years. In this role, he championed how data can be generated and used to drive structure- and machine learning-based design of reagents, biocatalysts, and therapeutic proteins to support new ways of working in discovery and development. Now leading the newly formed Data and Predictive Sciences organization that centralizes data science, computational design, and enterprise systems and data standards, Joel’s team enables a smooth flow of quality, model-ready experimental and in silico data for re-use across GSK R&D.

Seminars:

Implementing Active Learning Loops to Combat Bias & Improve Model Performance Over Time 2:40 pm

Discover how active and continuous learning systems allow models to improve with each new dataset, reducing initial bias Examine the impact of bias from datasets favoring only “good” molecules and how this limits model creativity and exploration Understand how integrating real-time lab results into training loops supports model refinement and validationRead more

day: Conference Day One

09:00 am | Workshop A | Lab-in-the-Loop AI Validation Systems for Improving Accuracy & Developability in De Novo Design 9:00 am

When experimental results are inconsistent, siloed, or simply too limited, how can we meaningfully train or refine AI models for biologics? Without standardized benchmarks or validation metrics, how do we know which models perform best, or even perform reliably at all? And if lab feedback is critical, what happens when infrastructure gaps, slow turnaround times,…Read more

day: Pre-Conference AI Validation Workshop Day