Overview
PMX is a predictive maintenance AI platform built for defense aviation operations. It ingests legacy maintenance data from military logistics databases and applies machine learning to forecast part failures, enabling proactive maintenance scheduling and optimized parts procurement.
Key Features
- ▸ML models using survival analysis and time-series forecasting to predict part failures
- ▸Data ingestion pipelines pulling from military logistics databases (NALCOMIS, DECKPLATE)
- ▸Model serving architecture for real-time predictions
- ▸Leadership dashboards for maintenance decision-making
- ▸FedRAMP and IL4/IL5 compliance on Azure Government
- ▸RMF/eMASS security assessment for IATT authorization
Problem Statement
Military aircraft maintenance is largely reactive, leading to unplanned downtime and reduced fleet readiness. Legacy maintenance data from systems like NALCOMIS and DECKPLATE exists but isn't being leveraged for predictive insights.
Leadership needed data-driven forecasting to optimize maintenance scheduling and parts procurement, shifting from a reactive to a proactive maintenance posture.
Technical Approach
Built on Azure Government cloud infrastructure with data ingestion pipelines pulling from military logistics databases. The platform applies feature engineering to raw maintenance records, then trains ML models using survival analysis and time-series forecasting techniques to predict component failure timelines.
The model serving architecture delivers real-time predictions to a dashboard layer, giving leadership actionable insights into fleet maintenance needs. All infrastructure operates within FedRAMP and IL4/IL5 compliance boundaries.
Architecture Overview
The platform follows a standard ML pipeline architecture: data sources feed into ingestion/ETL processes, which flow through feature engineering into ML model training. Trained models are deployed through a serving layer that powers visualization dashboards for leadership consumption. The entire stack runs on Azure Government with appropriate security controls.
Outcomes & Impact
Achieved a 36% improvement in prediction accuracy over baseline methods, significantly enhancing the reliability of maintenance forecasting.
Platform briefed to General and Colonel-level military leadership (2D MAW Aviation Logistics Department) and is actively used for maintenance decision-making across the organization.
Successfully navigated the RMF/eMASS security assessment process to obtain an Interim Authority to Test (IATT).
Technologies Used
Project Info
- Category
- AI/ML Platform
- Year
- 2025-2026
- Status
- Active
- Role
- Principal AI/Cloud Engineer