The Defense Innovation Unit wants industry to build AI tools that merge live video, satellite imagery, radar, sensor feeds, geospatial data and classified intelligence reports into one operational picture of space and missile threats.
The solicitation, reported by ExecutiveGov, sets aggressive timing. Data must reach a display or a query within five seconds, with a two-second objective. Systems should absorb normal flows of 20 to 30 megabytes per minute and manage bursts approaching 5 gigabytes per minute.
Reading a crowded sky
Attribution matters as much as detection. The department wants software that can name a threat rather than merely spot it, and that raises high-confidence alerts when objects fly close together or share an orbital neighborhood, including when sensor data thins out.
Requested techniques include machine learning, computer vision, predictive analytics, semantic technologies and knowledge graphs. Output has to work both as human-readable visualization and as machine-to-machine interfaces that plug into command-and-control workflows.
The security load
On the engineering side, designs must be cloud-native and containerized, follow NIST SP 800-171, and layer access controls for operations that involve several partners. Classified material and controlled unclassified information both need protection, and several years of history must stay searchable for after-the-fact analysis.
Bidders need a Facility Security Clearance and personnel cleared to the Top Secret level for the life of the agreement. Responses are due Sept. 24, with demonstrations run in an unclassified environment.
DIU is running the effort as a commercial solutions opening. A prototype that succeeds can become a production agreement without a further competition, which is why the short runway has drawn attention across the defense software market.
