Patent Portfolio Strategy
buildnyc.ai

A computer-implemented method for computing union labor penetration rates at the project, building, submarket, and jurisdictional levels using a two-layer attribution engine with a conditional co-occurrence rule. Layer 1 — anchor trade matching: contractor membership rosters for three specific trades — IBEW (electrical), Plumbers (UA), and Sprinkler Fitters (UA) — are scraped from union websites and cross-referenced against contractor entities on permit filings. These three trades are selected as anchors because their contractors are identifiable in permit records and their rosters are programmatically accessible; matched entities receive confirmed union attribution. A union GC registration — the general contractor entity matched against union GC membership records — serves as an additional co-occurrence kicker: a union-registered GC on the project strengthens the inference that anchor-confirmed union conditions apply to the full project team. Layer 2 — co-occurrence inference: when anchor trade contractors are confirmed union on a project (and no anchor trade is confirmed non-union), all other trades on that same project are inferred union via co-occurrence. Hard dissolve rule: if even one anchor trade on a project is confirmed non-union, co-occurrence inference is suppressed for the entire project — the project is classified open shop regardless of GC registration status. Union attribution is aggregated across all active projects within a defined market — by borough, work type, or jurisdiction — to derive an overall market penetration rate. Output: a statistically defensible union penetration percentage for a defined market, formatted for use in administrative proceedings, litigation, and prevailing wage coverage determinations.

A method for computing a construction contractor's current utilization, active backlog, and forward workload from public records — without contractor cooperation. Three permit states are aligned: backlog (filed, not yet permitted), active (PW2 issued, not expired), and runoff (permitted, nearing expiration). Labor hours are estimated via a dual-source equipment extraction model: (1) DOB NOW permit pages directly expose equipment types and quantities in structured fields — sprinkler heads, plumbing fixtures, risers, meters, electrical switches, panels, and other trade-specific items — scraped programmatically at scale; (2) plan document OCR extracts the same quantities from architectural drawings as a supplementary and validation layer. Labor hours are estimated from a synthesis of all available equipment quantities per trade — per-unit rates (NECA manhour tables, fixture unit standards, trade-specific coefficients) applied across the full inventory to produce composite project-level labor hour estimates. Aggregated output: contractor utilization score, forward workload in labor hours, and growth trajectory — derived entirely from public sources.

A computer-implemented method for grouping individual permit filings from multiple government systems into unified project representations with complexity classification. Combines cross-system deduplication (BIS + DOB NOW), job type hierarchy (NB > A1 > DM), and filing thread sequencing (I1→P1→P2 = same project thread) with a 5-dimension complexity scoring formula: scope×2 + cost×3 + trade count×2 + floor area×2 + structural bonus (max 44). Outputs first-class Project nodes classified into MINOR / STANDARD / MAJOR / LARGE / MEGA tiers.

A system for predicting regulatory objections to construction permit applications before submission. An OCR pipeline extracts text and metadata from digitized architectural plan drawings. A scraper collects historical objection records linking objection code + examiner identity + plan filing. A prediction model trained on these linked records outputs — for a given examiner × work type × code section × drawing metadata — the probability that a specific objection will be raised. Enables pre-submission plan review against individual examiner behavioral patterns.

A multi-signal algorithm for ranking which documents within a large government document corpus to process first, maximizing commercial value extracted per unit of processing time. Signals combined: permit issuance status (PW2), objection linkage, job filing recency, document discipline type, and filing sequence position (initial vs. revision). Output: a dynamic priority queue that routes high-value documents to processing resources before low-value documents — cutting time-to-value from years to weeks on corpora of 500k+ documents.

A probabilistic method for classifying whether a building is actively under construction using public permit record signals exclusively. Five signal categories combined: PW2 issuance status, ECB violation recency and type, inspection record dates, permit expiration proximity, and filing type pattern (NB vs. A1 vs. DM). The specific signal selection — particularly PW2 as the primary gate rather than any permit filing — and the construction-domain weighting are the inventive steps. Output: a building activity score with confidence interval used for underwriting, investment screening, and market intelligence.