Blinkly ODIN (Open-Source Data Integration for Navigation of the battlespace)

Contemporary military planning is constrained by the time, labor, and subjectivity required to transform unstructured open-source data into actionable Intelligence Preparation of the Battlespace (IPB) products. Frameworks such as the Marine Corps Planning Process (MCPP) and Joint doctrine depend on timely IPB to inform force disposition, adversary intent, and civil-terrain overlays. Current workflows are manual, error-prone, and brittle under data overload or adversarial manipulation. Prior automation attempts lack doctrinal fidelity, symbolic reasoning, and calibrated trust modelsŚlimiting their utility for real-world mission planning.Blinkly ODIN introduces an AI-native platform that autonomously extracts, evaluates, and synthesizes heterogeneous open-source inputs into machine-readable, doctrinally aligned IPB artifacts. ODIN integrates three core components: (1) a schema-constrained COA-GPT transformer for generating structured IPB products, (2) a symbolic reasoning engine for adversary COA extraction, and (3) a Bayesian trust model for source-level reliability estimation. The system ingests multi-format OSINTŚincluding HTML, PDF, GeoTIFF, and JSONŚand outputs structured adversary orders of battle, civil overlays, and NetCDF mobility grids. Each artifact is annotated with uncertainty scores and cryptographic provenance for mission transparency and downstream validation.On adversarially perturbed OSINT corpora, ODIN achieves ?90% alignment with SME-generated IPB products, ?5% hallucination drift, and ?0.75 correlation between Bayesian trust scores and human-rated source accuracy. COA-GPT produces outputs for 20-document clusters in under 60 seconds. The symbolic planner constructs multistep adversary narratives with spatial and temporal consistency, while the trust engine filters misinformation with ?88% precision. All outputs are compatible with automated post-hoc auditing and doctrinal red-teaming.Phase I will deliver a validated prototype with containerized ingestion, reasoning, and generation modules, a command-line interface, and modular APIs for integration with C2ISR environments. A reproducibility harness, doctrinal testbed of red-teamed planning scenarios, and evaluation protocol mapped to MCPP and IPB Annex C structure will be included. All components will conform to CUI restrictions and DoDI 3000.09, NIST RMF, and DoD AI Assurance guidelines for trusted autonomy.By uniting transformer-based schema generation, symbolic inference, and probabilistic trust modeling, ODIN establishes a secure, scalable, and mission-aligned foundation for IPB automation from open-source intelligence. It directly supports JADC2, expeditionary planning, and contested-domain intelligence fusion under emerging multi-domain operational requirements.