Dan Whitenack

Dan Whitenack, also known as Daniel Whitenack and Data Dan, is the founder and CEO of Prediction Guard and a co-host of Practical AI. His work includes deploying machine-learning systems, coordinating research for low-resource languages, and advocating private AI infrastructure for organizations handling sensitive data.[2][3][4][6]

Whitenack frames Prediction Guard’s purpose around privacy and institutional trust. In its May 2025 announcement, he said: “We’re on a mission to restore trust in human institutions via private, secure intelligence.” The company’s stated approach was to bring AI behind customers’ firewalls, allowing developers to build while meeting security and compliance requirements.[6]

In May 2025, Prediction Guard planned to continue developing deployment and governance capabilities for on-premises, air-gapped, and cloud VPC environments.[6]

Whitenack co-hosts Practical AI. In the July 2023 crossover with Latent Space, he joined a discussion of the podcast’s origins, lessons from earlier episodes, emerging AI trends, and his interest in low-resource applications. The introduction described him as having co-hosted Practical AI for the preceding five years.[3][1]

His personal biography describes more than ten years of developing and deploying machine-learning models at scale and building data teams at two startups and an international NGO. He also occasionally teaches data science and analytics at Purdue University.[2]

Before the 2023 crossover and his later public statements about Prediction Guard, Whitenack documented a different deployment challenge at SIL International: coordinating NLP researchers, datasets, and compute across countries.[1][6][4][5]

In his October 14, 2021 account, Whitenack described working as a data scientist at SIL International on low-resource and multilingual NLP. The team’s applications included spoken language identification, multilingual dialogue systems, machine translation, and translation quality estimation. He recalled starting as SIL’s only data scientist; by that account, its AI and NLP teams included ten or more data scientists or computational linguists.[4][5]

The researchers worked alongside technologists, linguists, translators, and university and industry partners in different locations. To coordinate their work, the team adopted ClearML for experiment tracking and shared compute management while retaining Google Colab, private GPU workstations, shared GPUs, and existing storage tools. ClearML registered experiments centrally, tracked data lineage and model provenance, and managed GPU queues. Unique dataset identifiers linked data to experiments, helping collaborators compare and reuse work across the infrastructure.[4]

Whitenack illustrated the resulting workflow with an Asia-based researcher prototyping training in Colab and registering the experiment and dataset. A collaborator in Europe could then clone that experiment, change the dataset and hyperparameters, and queue a run on the shared GPU server in Dallas. This was an example of what the infrastructure enabled, rather than an account of a particular completed experiment.[4]