When the U.S. military wants to know which AI tools can survive real-world combat conditions, it doesn’t issue a white paper—it runs a bake-off.
That’s exactly what happened last fall, when U.S. Central Command (CENTCOM), in partnership with the Pentagon’s Chief Digital and Artificial Intelligence Office (CDAO), hosted a five-day vendor evaluation to test how quickly companies could operationalize AI at the tactical edge. The goal wasn’t theoretical innovation. It was speed, reliability, and usefulness under pressure.
Raft, a defense technology company focused on software-defined warfare, emerged as the winner. The company was awarded a highly competitive Other Transaction Agreement (OTA), a contracting vehicle reserved for technologies that demonstrate clear operational advantage without the friction of traditional procurement.
The takeaway: AI for the battlefield is shifting from centralized labs and data science teams to the operators themselves—and Raft’s platform is built for exactly that transition.
From AI Experiments to Operator-Owned Capability
At the center of Raft’s win is its AI Mission System, known as [R] AIMS, a no-code, agentic AI platform designed to let military operators build, train, and deploy computer vision models without needing a data science background.
That’s a notable departure from how military AI has traditionally been deployed. Most computer vision systems used in intelligence, surveillance, and reconnaissance (ISR) rely on centralized model development, long update cycles, and specialized technical teams. That model works—until the mission changes faster than the software.
Raft’s pitch to CENTCOM was straightforward: give operators the ability to create and adapt AI models on the fly, directly in the operational environment, without waiting weeks or months for updates.
During the bake-off, Raft demonstrated a fully containerized platform that allowed users to train, evaluate, and deploy new models with built-in confidence scoring and traceability. The system handled real-world formats like NITF and supported live object tracking and optical threat detection—capabilities that typically require significant backend support.
In short, it worked where it mattered.
Why No-Code AI Matters in Combat Zones
The rise of no-code and low-code platforms has already reshaped enterprise IT, allowing business users to automate workflows without developers. Raft is applying the same principle to military AI, where the stakes—and constraints—are much higher.
In contested environments, connectivity is unreliable, missions evolve rapidly, and waiting for centralized updates can mean missed opportunities or increased risk. By removing the dependency on data scientists and centralized AI teams, [R] AIMS lets operators respond to unplanned ISR requirements in near real time.
That capability proved critical during the CENTCOM evaluation, where teams were tasked with adapting to changing scenarios over the course of the event. Raft’s system enabled rapid “gap-filler” model creation—custom computer vision models built specifically for emerging mission needs.
This isn’t just about speed. It’s about ownership. When operators understand and control the AI they’re using, trust increases, feedback loops tighten, and adoption accelerates.
Built to Plug Into Existing Military Workflows
One reason Raft’s platform stood out is that it doesn’t ask the military to rip and replace existing systems. [R] AIMS integrates directly with NGA Maven workflows, a cornerstone of the Department of Defense’s computer vision efforts.
That integration allows units to refresh models in days rather than quarters, while still maintaining compatibility with established intelligence pipelines. For combatant commands juggling legacy systems and modern AI initiatives, that kind of interoperability is often the difference between deployment and pilot purgatory.
During the bake-off, Raft highlighted several high-impact use cases:
- Satellite Broad Area Search: Operators can quickly train models to scan overhead imagery for mission-specific targets, filling gaps when existing models fall short.
- Distributed Maritime Monitoring: Task forces such as TF-59 can deploy and adapt computer vision models across complex maritime environments without centralized coordination.
- Counter-UxS Threat Detection: Units can build and update models to detect and classify optical threats—like unmanned systems—without waiting for enterprise-level updates.
Each use case reflects a broader shift in military AI: away from static, one-size-fits-all models and toward adaptable, mission-owned capabilities.
Agentic AI With Guardrails Built In
While “agentic AI” has become a buzzword in commercial tech—often associated with autonomous agents that plan and act independently—Raft applies the concept in a tightly controlled, mission-aligned way.
[R] AIMS is designed around machine-assisted feedback loops and built-in guardrails for Responsible AI. Every model trained on the platform is traceable, auditable, and aligned with mission intent. That matters in defense contexts, where explainability and accountability aren’t optional.
Instead of opaque black-box models, operators can see how models perform, understand their limitations, and adjust them as conditions change. The result is autonomy without loss of oversight—a balance the Department of Defense has been actively seeking as AI adoption accelerates.
As Raft CTO Bhaarat Sharma put it, the real breakthrough isn’t autonomy for autonomy’s sake. It’s enabling combatant commands to “build their own AI at the speed of the mission.”
The Competitive Landscape: Why Raft Stood Out
The defense AI market is crowded, with incumbents, hyperscalers, and well-funded startups all vying for relevance. Many offer powerful models, but few prioritize usability at the operator level.
What differentiates Raft is its focus on deployment over demonstration. Rather than showcasing impressive algorithms in controlled environments, the company emphasizes tools that can survive bandwidth constraints, data variability, and operational uncertainty.
That focus aligns closely with recent CDAO and CENTCOM priorities, which increasingly emphasize field-ready AI over experimental prototypes. The bake-off format itself reflects that shift, favoring solutions that can be used immediately rather than refined indefinitely.
Raft’s OTA win signals that the Pentagon is serious about decentralizing AI development—and rewarding vendors that enable that shift.
Implications for the Future of Military AI
[R] AIMS represents more than a single contract win. It points to a broader evolution in how AI will be developed, deployed, and governed across the U.S. military.
As conflicts become more data-driven and contested environments more dynamic, the ability to adapt intelligence systems in real time becomes a strategic advantage. Platforms that put AI creation directly in the hands of operators could shorten intelligence cycles, improve situational awareness, and reduce reliance on centralized bottlenecks.
Raft is now positioned to expand [R] AIMS across additional combatant commands, potentially setting a new baseline for how computer vision and agentic AI are used at the tactical edge.
For defense tech watchers, the message is clear: the next phase of military AI won’t be defined by bigger models or flashier demos, but by who can deliver usable, adaptable tools where decisions are actually made.
And in that contest, Raft just made a compelling case.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
