Applied AI for utility-scale solar. Engramiq ingests heterogeneous engineering, operational, and contractual data — single-line electrical diagrams, site-layout plans, datasheets, inspection reports, telemetry — and turns a fragmented body of documentation into a coherent, queryable representation of each asset.
It surfaces compliance issues, performance anomalies, and contractual exposures that would otherwise require manual review by senior engineers — and it's built on the same primitives that govern technical drawing review: geometry extraction, coordinate reconciliation, rule-based geometric assertion, and citation-anchored findings.
Computer vision, OCR, and CAD parsers (PDF / DXF / DWG) reconciled into a normalised geometric model for clearance, exclusion-zone, and consistency checks.
A domain-specific training pipeline — supervised, weakly-supervised, and synthetic augmentation — recognising panels, inverters, combiners, transformers, and electrical symbols.
Retrieval-augmented generation with structured extraction — every finding traceable to a source span and a governing clause. No finding without a citation.
Telemetry cross-referenced against single-line diagrams and as-built plans — surfacing inconsistencies across sources, the same shape as cross-discipline conflict detection.
In production: 5,005 components extracted & categorised · 58 site blocks geospatially verified against satellite imagery · human-in-the-loop QA where reviewers hold final authority. engramiq.com →
Domain-specific drawing intelligence with grounded findings and a human in command.
Start a conversation →
YOBI