Product family
The Workstation and Intelligence Engine are designed to run on a GPU machine you own. Your watchlists, holdings, notes, and questions stay inside your network.
The product is in development. This page describes the system we are building. It is not a description of a shipped release, and no performance or accuracy claim on this site has been measured yet. No software license or hardware is offered for sale on this site.
The investigation interface. An alert inbox, entity dossiers, a relationship graph, timelines, a filing viewer with year-over-year 10-K comparison, cited answers to plain-English questions, and case files you can export.
Free. Without an engine license, it covers the AI sector universe with one case file and watermarked exports.
The closed engine behind the Workstation: entity resolution across SEC, LEI, sanctions, and registry data; the relationship graph; rule-based red-flag signals; corroboration and confidence grading; watchlist monitoring and contagion alerts.
Annual license, priced by hardware tier.
Sector-wide processing we run centrally and deliver as signed, versioned snapshots. The AI capex chain pack is the first one and will be free. A customer who does not want to reveal even which sector they follow can run the whole pipeline locally.
Deployment, training, custom public-data sources, and custom analytics. Useful for institutions that need a tailored build or an air-gapped install.
The software is designed to run on a wide range of CUDA-capable devices, including consumer-grade cards, as well as CPUs. Performance varies by hardware. We do not sell, recommend, or assist in procuring hardware. Reference hardware profiles. Export compliance statement.
Every conclusion shows its work
Every factual statement is designed to link to the filing passage it came from. Confidence grades describe the evidence, not the company.
Each entity attribute, relationship, event, and signal links to a document, section, character offsets, and a content hash.
A source-reliability letter and a credibility number. Grading is deterministic code, never a language model.
GDLG does not review individual items. The customer's analysts review, annotate, and decide inside the Workstation.
Boundaries
Deployment
A hardware profile decides model size, quantization, batch size, and concurrency. Bigger machines serve more people, run more model replicas, and recompute whole sectors locally.