Skip to content
Agilevent

Yale University

We developed Kamino with Yale — a secure AI research platform that puts 12+ billion clinical records and on-demand GPU compute in researchers' hands.

Yale University logo
Industry
Higher Education & Health Research
Engagement
Platform Engineering · AI Infrastructure · Data Engineering
Website
yale.edu

With teams from Yale School of Medicine and Yale New Haven Health System, Agilevent developed Kamino — a computational health platform that lets medical AI researchers work with enormous amounts of real-world clinical data without compromising security, compliance, or speed.

The problem Kamino solves is one every research hospital faces: electronic health records are the richest raw material in medicine — Yale New Haven Health’s standardized repository holds more than 12 billion records — but putting them to work for AI research means navigating IRB compliance, extracting huge datasets safely, and provisioning serious GPU compute. The existing tools weren’t built for that. Kamino is.

The platform gives every research team three things:

  • IRB-aligned team workspaces — role-based access control built around the approved study protocol, so collaboration is secure by construction.
  • An automated ETL pipeline — researchers request the cohorts and data elements they need from the standardized OMOP data model, and the platform extracts, transforms, and delivers the data into their environment with read-only integrity guarantees.
  • On-demand AI compute — containerized environments with researcher-chosen CPU, GPU, and memory configurations, orchestrated on Kubernetes and ready for deep-learning frameworks out of the box.

It performs: in the published benchmark, a clinical natural-language-processing task using a 7-billion-parameter LLM scaled from 1,740 tokens per minute on a single GPU to 12,360 tokens per minute across eight GPUs — a 7.1× speedup — and the platform ingests real-time clinical data streams that have generated over 700 million data points. Kamino has supported COVID-19 outcomes research, next-generation phenotyping, and real-world evidence studies for biomedical devices and blood products.

The work is peer-reviewed: Kamino: A Scalable Architecture to Support Medical AI Research Using Large Real World Data (IEEE), co-authored by Agilevent’s Donn Felker with the Yale team.

Secure data platforms and AI infrastructure at institutional scale — that’s our lane. Get in touch.