How OpenEvidence built a healthcare AI that physicians actually trust
By Steven Van ·
OpenEvidence's frontend, built on Next.js and Vercel, held steady through a TikTok-driven traffic spike and cut serverless spend 90% with Fluid compute.
When a TikTok video sent traffic to Vercel-hosted OpenEvidence, according to a case study Vercel published. Response times stayed fast and error rates stayed near zero, with no manual scaling or extra provisioning needed.
OpenEvidence is a clinical decision support platform used daily by more than 40% of physicians in the United States, across over 10,000 hospitals and medical centers, and it supported more than 20 million clinical consultations in January 2026. Its backend runs on Python and Google Cloud Platform, while its frontend is built with Next.js and deployed on Vercel. Lead Frontend Engineer Andy Yoon says he's "pretty much the only engineer" on the team with a frontend background, with most colleagues working in Python and machine learning. Each commit deploys automatically, production deploys take five minutes, and every branch gets a preview URL, which the team has also used to spin up early proof-of-concept projects on their own custom domains for demos and enterprise partnerships.
After turning on Vercel's Fluid compute, which combines on-demand execution with server-like efficiency, OpenEvidence's serverless spend dropped by 90%. VP of Engineering Micah Smith said Vercel now makes up less than 5% of the company's overall infrastructure spend, even after 1000x growth.
