Supporting article
Download the published AutoDSM article for background, methodology, and validation details.
AutoDSM uses large language models and retrieval-augmented generation to identify system components and dependencies from documents, helping teams draft Design Structure Matrices faster with less manual effort.
Reproduced 357 of 462 DSM entries in the reported diesel engine validation study.
Prototype runtime achieved on a standard laptop-class machine.
Designed to support practical testing, validation, and future workflow development.
DSM creation is often manual, interview-heavy, and slow. AutoDSM is positioned as a practical front-end workflow that helps you get to a draft matrix faster, using your document base as the starting point.
Start from existing reports, manuals, and technical documents instead of building the matrix from scratch.
Ground outputs in proprietary data so results can reflect your own naming, scope, and system boundaries.
AutoDSM is designed to identify likely system elements and map likely relationships into a first-pass DSM.
The output is most useful as a reviewable starting point that experts can validate, refine, and complete.
AutoDSM breaks source documents into searchable chunks, retrieves relevant passages, then uses an LLM to propose system elements and dependency entries for the matrix.
Technical documents are split and embedded so relevant passages can be retrieved later.
Structured prompts query the retrieved content to find major components and likely dependencies.
The outputs are assembled into a matrix that can be inspected, refined, and used as a practical starting point.
Download the published AutoDSM article for background, methodology, and validation details.
Want to use AutoDSM or discuss your workflow? Reach out by email and we can continue from there.