cortex-recon

To inventory ML assets, you need to catalog models, pipelines, data sources, and monitoring configurations across projects. This process captures metadata like model versions, serving methods, schedules, data lineage, and cost signals to establish a single source of truth.

ML reconnaissance is the process of unifying and cataloging all machine learning assets to eliminate hidden silos and misconfigurations. You need it when asked for an ML assessment or to generate an up-to-date inventory and health snapshot of your deployed models.

You can identify deployed models and training pipelines by capturing metadata across projects and environments. This includes tracking model versions, serving methods, data inputs, feature stores, and schedules to ensure full auditability of your ML infrastructure.

Yes, you can catalog data lineage and cost signals for ML pipelines by capturing metadata across data feeds, feature stores, and monitoring configurations. This enables comprehensive evaluation and auditability of all data inputs and associated costs.

Yes, ML asset monitoring operates without external dependencies to inventory models and pipelines across multiple environments. It captures configurations, schedules, and data sources to reduce hidden silos regardless of your project setup.

The best way to generate an ML health snapshot and reduce hidden silos is running a comprehensive reconnaissance process. This inventories all assets, captures metadata for auditability, and creates a single source of truth across all projects.