Construction Ontology
A directed acyclic graph of construction work — 91 task codes across 13 phases, built from 269 days of real field work by ML Systems founder Sal Parvez. Not simulated. Not AI-generated. Ground truth.
81
Construction Codes
34
Deconstruction Codes
13
Phases
36
Primitives
269
Days Recorded
3 Neural Nets
Physical Net
Ground truth from construction sites → measurement and analysis → robot execution parameters
- ML1 (Manual Labor)
- ML2 (Measured Learning)
- ML3 (Machine Learning)
Financial Net
Market intelligence → capital structure → operational accounting
- LM (Language Modeler)
- FA (Financial Architect)
- AE (Accounting Engineer)
Dance Net
Community engagement → creative expression → long-term wealth building
- EV (Events)
- CR (Creativity)
- GW (Generational Wealth)
13-Phase Construction Sequence
Site Preparation
5 task codes
Foundation
7 task codes
Framing — Floor
6 task codes
Framing — Walls
8 task codes
Framing — Roof
7 task codes
Sheathing & WRB
5 task codes
Mechanical Rough-In
9 task codes
Insulation & Drywall
6 task codes
Exterior Finishes
7 task codes
Interior Finishes
8 task codes
Trim & Millwork
5 task codes
Fixtures & Hardware
4 task codes
Punchlist & Closeout
4 task codes
Robot Deployment Roadmap
Near-Term
8 codes
Repetitive, low-precision: sheathing, insulation, basic framing
Medium-Term
23 codes
Moderate precision: site prep, floor/roof framing, exterior finishes
Long-Term
24 codes
High precision: foundation, interior finishes, trim, fixtures
Human-Essential
13 codes
Judgment-intensive: mechanical, punchlist, closeout, inspections
The Lucent Lens
Glow within to help humans. The ontology exists to augment human capability, not replace it. Tier 4 tasks remain human-essential by design. Every automation decision passes through the Lucent Lens: does this help the person, or just the process?
Full ontology data available via the Data API at Enterprise tier. Includes task codes, sequence DAG, primitives, ML1 metadata, and robot parameters.