Case Study: OpenEvidence achieves faster, more cost-effective medical model training with Baseten Training

OpenEvidence trains medical models 23x faster with Baseten

OpenEvidence, a medical AI company providing an AI-powered search platform for physicians, faced significant challenges in training their custom, domain-specific large language models. They needed to rapidly experiment with training configurations to close the quality gap with closed-source models, but were bottlenecked by infrastructure that required sequential job execution, causing long delays. This slow iteration speed made it difficult to keep pace with new medical information and model releases.

Using Baseten Training, OpenEvidence gained the ability to run dozens of training jobs in parallel, allowing for extensive hyperparameter sweeps. The solution provided on-demand compute and a flexible, code-first environment for their machine learning engineers. As a result, Baseten helped OpenEvidence achieve a 23x improvement in latency, is projected to save them $1.9 million, and enabled even non-expert engineers to train powerful models in under 30 minutes.

OpenEvidence