Harell Data has entered into a multi-year agreement with CoreWeave, Inc. (Nasdaq: CRWV) to deploy its secure AI training platform on CoreWeave Cloud, according to a press release from CoreWeave.

Under the arrangement, training, fine-tuning, and inference workloads will run on Nvidia A100 and Nvidia Hopper GPUs within the Harell Data environment. The platform is structured so that raw data does not leave the Harell system during training runs; only the resulting model exits the platform.

Harell Data operates a marketplace model in which data owners receive a revenue share each time a training job runs against their dataset. Model builders can list trained models on the Harell marketplace under a Models as a Service arrangement and collect a fee each time the model is used.

The platform is aimed at biotech and biology research teams that work with proprietary scientific datasets. Harell Data was founded by Harlan Robins, PhD, co-founder of Adaptive Biotechnologies, and is headquartered in Bellevue, Washington, with an office in Palo Alto, California.

"High-quality, domain-specific data is absolutely a primary catalyst for the next generation of AI breakthroughs," said Robins. "We created Harell Data to connect researchers to proprietary scientific datasets for model training and inference without requiring them to view or extract the underlying data."

CoreWeave completed its public listing on Nasdaq in March 2025 and is based in Livingston, New Jersey.