From the source
OpenEvidence runs its medical search for 650,000 doctors on Vercel, and the platform's Fluid compute cut its serverless spend by 90% while performance held.
- The company's small engineering team, mostly Python and machine learning specialists, builds the physician-facing web application with Next.js and relies on Vercel for deployments and capacity.
- OpenEvidence supported more than 20 million clinical consultations in January 2026, and the company reported that more than 100 million Americans were treated by doctors using it in 2025.
- Engineers push to a development branch to get a preview link for demos, and merging to production triggers a build that reaches doctors in about five minutes, with rollback to a previous deployment if a release needs reverting.
- When the product went viral on TikTok, reaching around two million views in under a week, response times stayed fast and error rates remained near zero without extra capacity provisioning.
- VP of Engineering Micah Smith said the team reduced serverless spend by 90% while maintaining the same performance, and that Vercel is less than 5% of overall infrastructure spend even at 1000x growth.
- Take Home, the first patient-facing application, has its own project and domain so engineers can update the patient experience without deploying the other applications.
