LIVE DATA LOADING
FETCHING CURRENT REGISTRY DATA • PROJECT STATUS WILL UPDATE AUTOMATICALLY
SYNCING
#27

Agentic Shipping Classification Review

0.00score
Open GitHub RepositoryOpen Project WebsiteShare card
0
nigellauw/agentic-shipping-classification-review
OtherOPC-SG

roblem Statement Every commercial vessel must be certified by a classification society (e.g., DNV, ABS) before it can legally set sail. This multi-million dollar certification hangs on complex engineering schematics known as Piping and Instrumentation Diagrams (P&IDs). Currently, the "Class Review" process is plagued by three massive user pain points: The Bureaucratic Bottleneck: Shipyards must submit blueprints and wait 4 to 6 weeks for a human structural or marine engineer to manually cross-reference elements against thick binders of regulatory rulebooks. The Iterative Drain: The checking process is completely manual and opaque. Minor mistakes—like an invalid valve configuration or a missing safety redundancy on a cryogenic LNG line—result in a full rejection. Multiple submission-revision cycles are the norm, not the exception. Extreme Financial Risk: Large vessels cost upwards of a hundred million dollars to build. Every week of delay caused by slow, back-and-forth blueprint corrections eats directly into the shipowner’s capital return and pushes back shipyard build schedules. The Solution NaviRule is an agentic AI web-based workspace that acts as an automated, instant pre-submission gatekeeper. Instead of waiting over a month for external human audits, marine engineers upload their digital schematics directly into NaviRule. The system uses agentic AI to map out the blueprint's topology, evaluate it against codified regulatory sets in real time, and flag errors in seconds—transforming a months-long bottleneck into a lightning-fast internal clearinghouse. 2. Core Features (Hackathon Prototype MVP) To ensure a flawless live delivery within a strict hackathon timeline, the standalone web application prototype focuses exclusively on a single-file verification lifecycle using clean data: Feature 1: Web-Based File Upload Canvas (FR-1.1) A simple, browser-based drag-and-drop workspace where marine design engineers can upload their exported vector PDF blueprint schemas directly into the dashboard. Why it's hackathon-scoped: Restricting the MVP to vector PDFs/XML schema exports keeps data pristine for the AI, removing the complex computer vision/OCR risks associated with messy legacy scans. Feature 2: Interactive Web Layout Rendering (FR-1.2) Once uploaded, the web application renders a high-fidelity visual layout of the P&ID drawing natively inside the browser window. This provides the user with an intuitive, unified canvas to review their overall layout geometry. Feature 3: AI-Driven Rule Violation Mapping (FR-1.3) An agentic AI backend automatically cross-references the diagram's component topology against a codified subset of maritime laws (such as LNG fuel gas safety loops). Non-compliant sections (e.g., a missing fail-safe valve configuration) are dynamically highlighted in bright red or yellow bounding boxes directly overlaid on the blueprint components. Feature 4: Contextual Error & Actionable Remediation Side-Panel (FR-1.4) Clicking on any highlighted red or yellow violation box opens a dual side-panel interface. Powered by a specialized multi-agent pipeline, it displays a human-friendly compliance breakdown containing three fields: the exact nature of the violation, the specific binder rulebook index breached (e.g., DNV-RU-SHIP Pt.5 Ch.7), and concrete, actionable text instructions detailing how the engineer can physically modify the CAD layout to fix it.

Offline Participation: SG_OPC_0712

Score breakdown

GitHub size · 50%X / social · 50%
0.6×0+0.4×0=0
GitHub size0 (50%)
Stars0 → 0
Forks0 → 0
Size score0
Weight× 0.6
X / social0 (50%)
Likes0
Replies0
Quotes0
Reposts0
Social score0
Weight× 0.4

Category comparison

GitHubSocial

#3 / 84 in category · OtherTop 4%

This project0
Category avg0.63
Category best33.65

Score History

2026-07-25

0.00

Total Size Social
0.30.20.10.10.0Jul 12Jul 25

Execution flow

01Submit
GitHub repo + X post submitted with a wallet-signed payload.
02Fetch
Stars, forks and X engagement are pulled and snapshotted daily.
03Score
Size (GitHub) and social (X) are scored, then category-weighted.
04Rank
Ranked against the field; the daily trend is recorded.