U.S. Special Operations Command (SOCOM) is opening a new path for industry to plug into its drone autonomy stack: a hunt for a 'self-service' synthetic data generation platform that SOCOM operators can run themselves, with no vendor in the loop, to train computer vision models for unmanned systems.
The UxSAI Synthetic Data Modeler
The Unmanned Systems Autonomy and Interoperability (UxSAI) Synthetic Data Modeler assessment, hosted by SOCOM innovation hub SOFWERX, calls for a platform that produces labelled EO/IR training imagery for computer vision model development across air, ground and maritime unmanned systems. The catch is operational independence — SOCOM wants its personnel to operate the platform 'organically' inside the existing MLOps pipeline used by program offices and tactical squadrons.
Timeline: Whitepapers Due 29 June
Vendor whitepapers are due 29 June 2026, with a mid-July live demonstration event at which shortlisted companies will show their platforms in action. SOCOM will downselect in September and start contract negotiations with the chosen vendor or vendors in October 2026 — an unusually compressed timeline that mirrors how the command has been awarding fast-turn AI work.
Plugging Into A Wider Drone Push
The synthetic data effort feeds the same SOCOM autonomy stack supporting capabilities profiled in recent FPV procurement coverage and broader sweeping changes covered in the Senate NDAA push for a four-star autonomous systems combatant command.
Reporting based on coverage from DefenseScoop, GovConFeed, Inside Defense and SOFWERX events.