Skip to content

Plumbline.ai

Plumbline.ai is an AI co-pilot for construction lenders , banks and Construction Risk Management Consultants (CRMCs) verifying monthly draw requests against the plans they're financing.

Core trick

Compares photographed site evidence and drone imagery against sealed plan PDFs and G703 line items to issue a draw verdict.

Tags

Description

Plumbline.ai is an AI co-pilot for construction lenders , banks and Construction Risk Management Consultants (CRMCs) verifying monthly draw requests against the plans they're financing. It pairs frontier-grade reasoning (Claude Opus 4.7) with modern field telemetry , drones, IoT sensors , 360° walkthroughs, and fixed-position construction cameras, to turn an inspector's manual workflow into a structured, audit-ready pipeline. Today verification is manual: an inspector drives to site, eyeballs progress, and writes a narrative the loan officer takes on faith. OCC and FDIC have flagged weak Change Order and Schedule-of-Values discipline as a safety-and-soundness issue. An 8-agent Opus 4.7 architecture ingests sealed PDF plans across four disciplines (architectural, structural, electrical, plumbing), the loan's finance plan, the AIA G703 Schedule of Values, and authenticated site evidence. A Managed-Agents Supervisor then investigates findings autonomously and produces a Gap Report with a draw verdict — APPROVE / APPROVE_WITH_CONDITIONS / HOLD / REJECT, citing G703 line items and flagging unapproved scope. Two industries that resisted automation, bank credit risk and field construction , converge here. Agents do the cognition; drones and IoT bring the field witness. Built at the seam between regtech for financial services (my work) and 50 years of practice (my father's). MIT licensed
wutangasaf/hackathonopus★ 22 forksTypeScriptMITCreated Apr 22, 2026 · last push Apr 26, 2026

README excerpt

# Plumbline.ai

An AI co-pilot for construction lenders — banks and CRMCs verifying monthly draw requests against the approved plans they're financing. Plumbline.ai ingests the approved construction drawings and specifications (structured and unstructured — sealed PDF drawings, BoQ, AIA G703 Schedule of Values, CSI Masterformat specs), the loan's finance plan with milestone tranches, and authenticated site evidence (phone photos in the hackathon slice; drone, 360°, fixed-camera and IoT telemetry in the full product), then produces a Gap Report with a concrete draw verdict — `APPROVE`, `APPROVE_WITH_CONDITIONS`, `HOLD`, or `REJECT` — citing G703 SOV line items and flagging any unapproved scope deviations.

Built for Anthropic's *Built with Opus 4.7* hackathon (Apr 21–26 2026). MIT licensed.

## Why this exists

Construction lenders release loan tranches against milestone completion, but verification today is manual: an inspector drives to site, eyeballs progress, and writes a narrative. Regulators (OCC, FDIC) have repeatedly flagged inadequate Change Order tracking as a safety-and-soundness issue. Plumbline is modelled on 50 years of civil-engineering and Construction Risk Management Consultant (CRMC) practice and turns an inspector's workflow into a structured pipeline that cites evidence back to the approved documents.

## Product premise

Plumbline is built around three principles drawn straight from CRMC practice.

**1. Multi-source, chain-of-custody evidence.** A draw dec

Team