Proof-of-concept system on the aibomgen-v2/main branch for the next-generation AIBoMGen platform. It integrates with Kubeflow via an ML Metadata (MLMD) store and extracts full pipeline lineage to generate CycloneDX AI Bills of Materials (AIBOMs).
The system generates per-model and per-dataset BOMs, including lineage-aware relationships using BOM-Link URNs and explicit model–dataset dependencies via external references. It also provides an interactive graph-based viewer to explore pipelines, models, datasets, and their relationships.
The architecture consists of a Kubeflow-style MLMD stack with a simulator for pipeline execution, a BOM generation service, and a web-based visualization layer. This work is part of PhD research into end-to-end AI lifecycle traceability and supply chain transparency, conducted as part of the CRACY Project.