Generative AI in the Real World: Danielle Belgrave on Generative AI in Pharma and Medicine

Why It Matters

AI‑driven patient stratification and synthetic data generation promise faster, more precise drug development, while responsible‑AI frameworks are essential to mitigate risks in high‑stakes healthcare applications.

Key Takeaways

• AI accelerates patient stratification for clinical trial recruitment
• Generative models create synthetic multi‑omics data to speed drug discovery
• Responsible AI team enforces guardrails against hallucinations in health models
• Multimodal foundation models link genomics with computational pathology images
• Small‑sample robustness remains a major challenge for clinical AI

Summary

In this podcast, GSK’s Vice President of AI and Machine Learning, Danielle Belgrave, explains how generative and traditional AI are being deployed across the pharmaceutical pipeline. Drawing on a 15‑year career that spans a PhD on asthma heterogeneity, stints at Microsoft Research and DeepMind, and now leading AI for clinical development at GSK, she outlines the shift from one‑size‑fits‑all treatments to data‑driven, patient‑specific interventions. Belgrave highlights several use cases: leveraging whole‑genome and RNA‑seq data to inform computational pathology, using multimodal foundation models to translate molecular signals into biopsy image insights, and employing generative AI to synthesize large‑scale multi‑omics datasets for target identification. She also stresses the importance of responsible AI, noting GSK’s dedicated team that applies model cards, external reviews, and hallucination metrics to ensure safety and reproducibility. Concrete examples include her PhD work that identified five distinct asthma subtypes, current projects that map genomic profiles onto tissue‑level pathology, and internal language models such as “jewels” that assist scientific productivity while being rigorously vetted. The discussion underscores the breadth of data—clinical notes, biomarkers, microbiomes, epigenetics—and the technical hurdles of batch effects, small‑sample robustness, and data sparsity. The overarching implication is that AI, especially generative and multimodal models, can dramatically shorten drug discovery cycles, improve trial enrollment efficiency, and enable precision therapeutics, provided that robust governance and methodological safeguards keep pace with rapid innovation.

Original Description

Danielle Belgrave is VP of AI and machine learning at GSK (formerly GlaxoSmithKline). She joins Ben to discuss using AI and machine learning to get better diagnoses that reflect the differences between patients. Listen in to learn about the challenges of working with health data—a field where there’s both too much data and too little, and where hallucinations have serious consequences. And if you’re excited about healthcare, you’ll also find out how AI developers can get into the field.

Follow O'Reilly on:

LinkedIn: https://www.linkedin.com/company/oreilly/

Facebook: http://facebook.com/OReilly

Instagram: https://www.instagram.com/oreillymedia

BlueSky: https://bsky.app/profile/oreilly.bsky.social