NGA Awards ECS $104.3M "Canyonlands" Contract to Advance AI for GEOINT
The National Geospatial-Intelligence Agency awarded ECS the potential $104.3 million Canyonlands contract to develop and demonstrate AI capabilities. NGA disclosed the award on Sept. 19. The work centers on operational AI for geospatial intelligence across cloud and on-prem environments.
What Canyonlands Covers
Per a July notice on SAM.gov, Canyonlands is a sole-source award to ECS to transition NGA-funded research topic areas from an Army Research Laboratory-administered vehicle known as "Runway." The anticipated period of performance includes a 12-month base term as oversight moves from ARL to NGA. The scope focuses on turning research into fielded capability, not just proofs of concept.
SAM.gov and NGA provide official procurement and mission context.
Link to Project Runway
In 2023, the U.S. Army awarded ECS the $190 million Project Runway contract to build and demonstrate capabilities that provide AI mobility and agility at the field or theater level across multiple DOW AI domains. Runway also targeted GEOINT AI integration for both cloud-based and on-premises deployments. Under Canyonlands, ECS continues this work with NGA oversight, emphasizing transition and demonstration.
What This Means for IT and Development Teams
- MLOps across hybrid environments: consistent pipelines, model registries, and deployment patterns that work in cloud, on-prem, and disconnected settings.
- Model portability: containerized inference, export standards (e.g., ONNX), hardware abstraction, and operator toolchains that avoid vendor lock-in.
- Data engineering at GEOINT scale: tiling/streaming geospatial data, labeling workflows, synthetic data where needed, and lineage tracking.
- Security and compliance: RMF/ATO readiness, zero trust patterns, cross-domain solutions, SBOMs, and auditability for model changes.
- Evaluation and monitoring: mission-grounded metrics, bias/error analysis, adversarial testing, and drift detection for both cloud and edge deployments.
- Resource-aware inference: GPU/CPU planning, quantization, batching, and scheduling for constrained environments.
Expected Outputs to Watch
- Reference architectures for GEOINT AI across cloud and on-prem.
- Reusable MLOps templates: CI/CD for models, data versioning, and deployment blueprints.
- Testbeds and demo events that prove mission utility under realistic constraints.
- Model catalogs and datasets with clear provenance and evaluation reports.
How to Prepare If You're in the Ecosystem
- Package your AI services for hybrid: Kubernetes manifests, Helm charts, offline install options, and clear upgrade paths.
- Offer hardware flexibility: support common accelerators and configurations used in government environments.
- Ship compliance-ready artifacts: STIG guidance, SBOMs, security test results, and documentation mapped to NIST controls.
- Plan for disconnected ops: caching, local feature stores, model snapshotting, and no-callback licensing modes.
- Make evaluation easy: provide benchmark scripts, mission-relevant metrics, and red-team test sets.
Timeline and Oversight
Canyonlands begins with a 12-month base period under NGA oversight. Expect a fast cadence of demos and decision points tied to transition readiness and operational value.
Level Up Your Team
If your team supports federal AI delivery and needs to sharpen skills across MLOps, model deployment, and evaluation, explore role-based learning paths here: AI courses by job.
Your membership also unlocks:
