How it Works

GEOWatch leverages AI/ML to identify and classify satellite patterns of behavior (PoB) from passively collected ephemeris data. The system continuously updates its catalog with new inputs and detected PoB events, deepening insight into anomalous or concerning on-orbit activities and elevating items that merit operator review.

Mission use cases

  • Mission Analysis: Accelerate campaign and scenario planning with data-driven insights that reduce manual review and highlight what matters first. 

  • Intent Identification: Uncover adversary intent by analyzing behavioral patterns to support early warning and threat anticipation.

  • Satellite Characterization: Generate rapid, detailed profiles of space objects to support threat triage and mission assurance.

  • Anomaly Alerting: Deliver real-time alerts on unexpected behaviors so operators can act before anomalies escalate.

Integration & deployment

  • APIs & interoperability: Standards-based interfaces; drops into existing pipelines and tools used by SDA teams. 

  • Operational fit: Exercised with operators in the SDA TAP Lab; integrates with downstream systems for fusion, scoring, and reporting.

  • Environments: Cloud, edge, or air-gapped; supports multi-level security. 

Validation & Awards

  • Prototyped and exercised with government operators in SDA TAP Lab cohorts. 

  • MIT ARCLab (2024): recognition for satellite pattern-of-life identification. 

How it Works

SkySim produces synthetic, high-fidelity datasets and modeled scenarios that fuel predictive modeling, simulation, and mission readiness. Teams can rehearse workflows, evaluate policies, and refine decision support in a safe, repeatable environment prior to deployment.

Mission use cases

  • Dynamic Scenario Modeling: Accelerate testing and evaluation by exercising mission workflows and alert policies in realistic scenarios.

  • Synthetic Data Generation: Create scalable, mission-relevant datasets to improve model performance and analysis.

  • Operator Training & Readiness: Rehearse decision support and procedures before fielding to reduce risk at go-live.

How it Works

CoCoWatch provides launch-window analytics to support early threat awareness, using event-driven assessments to identify trajectories that could create coplanar alignment with high-value assets and reduce operational surprise.

Mission Use Case (proven in SDA TAP Lab)

  • Predictive Threat Awareness: Anticipates potential co-orbital threats before they occur, enhancing mission assurance. 

  • Mission-Phase Tailoring: Aligns workflows to each mission phase—from launch planning to on-orbit operations—for precision and relevance. 

  • Decision Advantage: Reduces risk and accelerates planning by identifying optimal launch windows and delivering actionable insights at mission speed. 

Validation & Awards

  • Prototyped and exercised with government operators in SDA TAP Lab cohorts. 

  • MIT ARCLab (2024): recognition for satellite pattern-of-life identification. 

How it Works

SDS acts as a centralized repository for cataloging key attributes of space objects. By aggregating open-source information and data from various providers, SDS facilitates the classification of satellites, aiding in tracking, analysis, and the development of behavior and signature models to identify indicators of camouflage, concealment, deception, and maneuver (CCDM).​

Mission use cases

  • Multi-Source Aggregation: Combine ephemeris, telemetry, characterization, and external datasets to enable richer cross-domain insights.

  • Accelerated Analysis: Reduce time to build comprehensive assessments by eliminating stovepipes and simplifying data integration.

  • Enhanced Mission Awareness: Provide a holistic picture of the space domain to support faster, more informed decision-making.

How it Works

Stonecutter is a secure AI/ML Battle Management engine that optimizes resources, enforces policy and lineage, and continuously evaluates and monitors models from lab to edge.

Mission use cases

  • Space Behavior Models: Learn “normal” behaviors for resident space objects (RSOs), detect deviations in near-real time, infer intent (e.g., benign inspection vs. hostile proximity),

    • Space behavior models learn baseline patterns for resident space objects and continuously score live tracks to detect anomalous maneuvers, providing early warning and evidence of intent.

    • They fuse OD/TLE/CDMs, RF/telemetry, photometry, and OSINT with Mod/Sim/Vis to replay hypotheses, forecast approach windows/keep-out breaches, and drive targeted multi-INT collections.

    • The result is signed Threat SITREPs with confidence metrics and COA recommendations (monitor, block, separation)—

    • If you’d like a brief case study with metrics and sample artifacts, reach out and we’ll share the details.

  • Predictive Space Weather: Anticipate geomagnetic/solar events (flares, CMEs, SEPs) and translate forecasts into concrete actions that protect links, navigation, and LEO constellation operations

    • Predictive space-weather models fuse solar/geomagnetic data to nowcast and forecast flares, CMEs, and SEPs—issuing early warnings with severity and arrival windows.

    • They translate forecasts into concrete mission impacts for SATCOM, GNSS, and LEO operations and auto-task mitigations (e.g., link adaptation, OD update cadence, drag makeup burns) to preserve availability.

    • We publish signed Space Weather SITREPs and live dashboards and close the loop with runtime monitoring

    • If you’d like a brief case study with metrics and sample artifacts, reach out and we’ll share the details.•

How it Works

Forge is a hardened ML/AI platform to develop, train, and run models with security-by-default, GPU-optimized performance, and edge-ready deployments from cloud to contested environments.

Mission use cases

  • Air-Gapped Retraining of a Space Behavior Model: Restore and harden behavior-of-life/intent models after drift (new adversary RPO patterns) entirely inside an IL5/IL6 air-gapped enclave, preserving provenance, security, and reproducibility

    • Keystone orchestrates an IL5/IL6 air-gapped retrain by importing curated datasets via CDS with full lineage, augmenting scenarios in Mod/Sim/Vis, and running reproducible GPU training and selection inside the enclave.

    • Standardized T&E produces audit-ready evidence (model cards, reports), then Forge containerizes and Stonecutter signs/attests a Deployment Promotion Package for canary and fleet rollout.

    • Post-deploy monitoring feeds back drift and operator input for continuous improvement.

    • If you’d like a brief case study with metrics and sample artifacts, reach out and we’ll share the details.

  • Deploying & Managing an ML Model on a Remote Sensor (Field or On-Orbit): Package, deliver, activate, and sustain an ML model on a constrained edge node (remote field sensor or spacecraft) with assured security, performance, and rollback.

    • Forge packages and optimizes the model for the target edge node (field or on-orbit), runs hardware-in-the-loop checks, and —together with Stonecutter — produces a signed Deployment Promotion Package with resource policies and telemetry.

    • Keystone then orchestrates staged delivery over constrained links and activates the model shadow → canary → fleet with watchdogs, quotas, and safe rollback, maintaining an auditable chain of custody.

    • We monitor performance and drift, convert detections into prioritized tips, and feed edge telemetry back for retrain and updates.

  • If you’d like a short case study with metrics and sample artifacts, reach out and we’ll share the details.