doc.ai: A mobile AI app blazes trails in medical research

doc.ai needed to scale ML training without heavy on-prem infrastructure while handling HIPAA-compliant health data for a fast-growing user base and training researchers/doctors in deep learning. It rebuilt on Google Cloud (Compute Engine GPUs, GKE, Cloud Storage, Cloud SQL), re-implementing its fast.ai platform from scratch in 2 days versus a week-plus on AWS. Batch experiment time dropped from two weeks to one day, GPU capacity scaled on demand from 100 to 500+, and doc.ai grew to 35,000+ active users in under a year (about 5,000 new users monthly); Head of AI Neeraj Kashyap said "You can move much faster building products on Google Cloud than on AWS" and called GKE "our most heavily used product."

Google supplies AI to pharma through several distinct lines — Gemini/Vertex AI as an enterprise platform, and AlphaFold-based drug design licensed commercially through Isomorphic Labs.

This sits in the Build stage of the AI use-case lifecycle: Named examples from the foundation-model labs and platforms pharma teams use to actually construct, train and generate the model or application.