With $106M contract, JAIC takes major step building central AI platform for DOD

The Joint Artificial Intelligence Center inked a $106 million contract with Deloitte Consulting to build out its Joint Common Foundation (JCF) AI development platform, which the center hopes will eventually bring a platform-as-a-service type capability to the Department of Defense.

The JCF is a cloud-enabled environment that has been developed in conjunction with the Air Force’s Platform One team for use across the DOD’s AI and machine learning initiatives. It’s a building block of the DOD’s AI toolkit as a single place coders can build models, work with data and develop AI systems.

Deloitte Consulting was selected by the JAIC, in partnership with the Defense Information Systems Agency, to “provide, operate, maintain, secure and enhance” the platform, according to the JAIC. Deloitte will act as an integrator with an array of sub-contractors working to develop JCF.

“This will allow us to quickly accelerate AI/Ml development across the DOD,” Col. Sang Han, chief of infrastructure and platform for the JCF, said in an interview.

So far, the Platform One team has been working with the JAIC to establish DevSecOps practices for the JCF. As Deloitte enters the fold, it will continue the security-at-every-level practices, Han said. The company will work with innovative small businesses while it handles the Pentagon’s bureaucratic acquisition processes to “lower the barriers” for the JAIC to bring commercial work to its tech stack.

‘Democratizing’ AI for DOD

The JCF is designed to be on unclassified and classified networks and provide users across the department an environment work with DOD data. Eventually, the hope is to have JCF be a marketplace of sorts for users across the services to work with tools developed in the environment by others.

“What we currently have is a bunch of disparate AI and ML development environments,” Han said. “Really, that doesn’t really help us rapidly adopt and scale AI across the DOD.” Rather, he said, the goal of the JCF is to unify the DOD’s AI work into one environment and “democratize” the use of AI.

JAIC’s vision for the future is that even someone in an austere environment like Afghanistan will be able to log in and design machine learning programs and work with DOD data to fit their mission needs. Eventually, the JCF will host an array of tools built by users across the services for a “one-click interface” to work with tools.

While it may be some time before these types of user experiences arrive, the contract spells out a timeline for scaling the JCF in the coming months. Initial milestones include delivering a toolset for building out the environment that can be sent through the “[authority to operate] ringer,” as Don Bitner, chief of strategy for the JCF, said in an interview.

“What we were really looking for is a single belly button that is going to be able to pull together both the commercial and open source tooling that is going to be required for AI and ML, as well as deploy it out across DOD,” Bitner said.

JEDI-ready

The system will be “interoperable and JEDI-ready” once the DOD finally gets its Joint Enterprise Defense Infrastructure cloud system, which is currently tied up in legal challenges, into operation. The CIO’s office, which houses the JAIC, anticipated that JEDI would be re-awarded this month. But the DOD this week asked for a further delay to its finalize its vendor selection for JEDI in the ongoing bid protest waged by Amazon Web Services against the Pentagon and contract winner Microsoft.

For now, the JCF is running on a provisionally authorized cloud.

“To the maximum extent possible, leveraging cloud service providers is going to be critical to making sure we can move data around,” Bitner said.

Correction: Aug. 13, 2020. The original story misstated Don Bitner’s name.