Case Study: AstraZeneca accelerates AI model training with Cerebras

AstraZeneca - Customer Case Study

AstraZeneca, a leading pharmaceutical company, needed a faster way to iterate and experiment with AI across hundreds of thousands of abstracts and research papers. Using Cerebras and its CS-1 system, the team aimed to accelerate model training and improve the speed and accuracy of research queries.

Cerebras implemented its CS-1 platform to train large BERT models far more efficiently than traditional GPU clusters. As a result, AstraZeneca reduced a training run that previously took more than two weeks to just over 2 days—52 hours on a single CS-1—enabling more frequent iteration and much faster, more accurate answers.

AstraZeneca

Nick Brown

Head of AI Engineering