ENDECTA
Contact Sales

Built for buildings

Carmkn is not a wrapper around an LLM. It is a spatial AI platform with a purpose-built runtime, deterministic engineering calculations, and a knowledge graph that learns from every project.
Intelligence

Carmkn — Learning Knowledge Graph

Carmkn builds a persistent knowledge graph from your project data where connections between documents, equipment, spaces, and specifications are preserved. The graph learns from every interaction — correct answers strengthen retrieval paths, wrong answers weaken them. It gets smarter with every project.
Adaptive learningRetrieval paths that consistently lead to correct answers are strengthened. Paths that produce wrong answers are weakened. The system actively avoids repeating mistakes and reinforces what works for your domain.
Pattern consolidationWhen the same successful approach appears across multiple projects and questions, Carmkn consolidates it into a reusable procedure. The knowledge graph grows its own intelligence based on what actually works.
Natural knowledge decayKnowledge that isn’t used fades over time, keeping the graph focused on what matters now. Frequently-used paths stay strong. Stale connections are cleaned up automatically while proven retrieval patterns are preserved.
Grounded responsesEvery answer is sourced from your actual project documents — not general AI training data. Responses include citations so you can verify exactly where the information came from.
Ingestion

Document Intelligence

Connect your project files from Egnyte, OneDrive, SharePoint, or upload directly. Carmkn processes 74+ file formats — including Revit models and IFC building information — and makes every entity, relationship, and specification searchable in plain English.
BIM model ingestionRevit models are ingested with complete object property extraction and spatial relationships preserved. IFC files (2X3, 4, 4X3) are mapped to equivalent categories. DWG files get text-level extraction. Levels, spaces, equipment, circuits, and supply chains become queryable nodes in the knowledge graph.
Professional PDF extractionTables, text, and structured content are extracted from PDFs with high-fidelity processing. Scanned documents get OCR processing. Images within documents can be analyzed visually for drawings, diagrams, and schedules.
Code and specification parsing45+ programming languages parsed with full structural awareness — functions, classes, and imports are preserved as semantic units, not just raw text. Construction specifications are parsed into structured entities with typed relationships.
Construction entity extraction50+ domain-specific entity types automatically identified — equipment, materials, specifications, NEC references, measurements. Entities are linked into the knowledge graph with typed relationships so you can traverse from a document to the equipment it describes to the circuits that serve it.
Engineering

Deterministic Calculations

The AI decides what to calculate. Verified code does the math. Every engineering calculation uses deterministic formulas with real NEC tables — never AI approximation. Calculations pull real panel, circuit, and fixture data from your ingested Revit models. Not hypothetical — actual project data.
NEC load calculationsFull Article 220 implementation with progressive demand factor tiers. Dwelling optional method per 220.82. Motor loads per 430.24. Non-coincident heating/cooling loads per 220.60. Results include NEC article citations to support review workflows.
Voltage drop analysisFull AC impedance formula using NEC Chapter 9 Table 9 values for all conductor types and conduit materials. Accounts for both resistance and reactance — critical for large conductors. Compliance check against NEC 210.19(A) and 215.2(A) thresholds.
Power flow & fault currentModels entire building electrical systems — transformers, feeders, panels, branch circuits. Newton-Raphson power flow analysis and IEC 60909 short-circuit calculations. Time-series analysis for daily, monthly, and annual load profiles.
Title 24 & CBECC complianceFull Title 24 lighting compliance: pulls fixture data from Revit, calculates actual lighting power density from fixture counts and wattages, compares against area category limits. CBECC energy code compliance analysis across 2019, 2022, and 2025 code cycles.

Supported file formats

Carmkn processes 74+ file formats out of the box. BIM models preserve spatial relationships. Office documents preserve tables. Code files preserve structure. Nothing is silently dropped.
Revit (.rvt)IFC (2X3/4/4X3)DWGPDFDOCXXLSXPPTXCSVImages (OCR + Vision)45+ code languagesEmail (.eml/.msg)HTMLMarkdownJSONEPUB
© 2024 Endecta LLC. All rights reserved. Carmkn can make mistakes so validate all responses. Carmkn executes with Claude (Anthropic).