Infleqtion lands $2M Army SBIR to build secured edge AI for PNT (SAPIENT)
Infleqtion won a $2 million Direct-to-Phase II SBIR from the U.S. Army to build SAPIENT-Secured AI for Positioning at the Edge, Navigation, and Timing. The 18-month effort targets resilient navigation and timing (PNT) in contested and degraded conditions. SAPIENT was selected out of 133 submissions through the xTechScalable AI Challenge, with full system work funded through October 2026.
What SAPIENT delivers
- Data fusion across GPS, inertial sensors, and computer vision with inference at the edge.
- Quantum-inspired multimodal learning and Boltzmann Machine-based models to keep estimates stable when signals are degraded or spoofed.
- Compact, power-efficient deployment for constrained platforms operating with unreliable connectivity.
Inside the stack
At the core is Infleqtion's Contextual Machine Learning (CML), built to learn from diverse data sources over extended time windows. CML is powered by NVIDIA's CUDA-Q platform and quantum-inspired algorithms, improving adaptability to shifting sensor quality and mission context.
The company has already applied CML in defense programs like QuIRC for RF signal processing, giving this effort a head start on real-world integration. The new contract funds full system development and testing, pushing the approach from prototype to fieldable capability.
Why this matters for IT and development teams
- Edge-first AI: Model architectures must hold up under tight compute, thermal, and power limits without cloud dependence.
- Sensor fusion engineering: Expect emphasis on cross-modal calibration, uncertainty modeling, and outlier handling.
- Security posture: Anti-spoof and adversarial resilience move from "nice to have" to baseline requirements.
- Lifecycle planning: The program runs through October 2026-plenty of time to align internal roadmaps for edge inference, testing, and validation.
Technical notes to consider
- Model compression and quantization strategies will be essential for compact hardware.
- Long-horizon context means careful handling of temporal drift, sensor dropouts, and state recovery.
- Boltzmann Machine components can help with structured uncertainty; watch for training stability and convergence constraints on-device.
- Test plans should simulate degraded GNSS, spoofing, and intermittent comms to validate resilience claims.
Infleqtion also plans to go public via a merger with Churchill Capital Corp X (NASDAQ: CCCX) in September 2025, which could expand resourcing for defense and dual-use deployments.
Sources and further reading
Business Wire: Army awards Infleqtion $2M for SAPIENT
NVIDIA CUDA Quantum (CUDA-Q)
Level up your edge AI skills
If you're building similar stacks-sensor fusion, on-device inference, and adversarial resilience-browse focused training paths by role:
Complete AI Training: Courses by job
Your membership also unlocks:
