Back
Explore EventsExplore ProjectsMy Projects
PRUDENCE (PRudential Urban Development Compliance Engine) — screenshot 1
PRUDENCE (PRudential Urban Development Compliance Engine) — screenshot 2
PRUDENCE (PRudential Urban Development Compliance Engine) — screenshot 3

PRUDENCE (PRudential Urban Development Compliance Engine)

India's AI Agent for Construction Compliance "Planned by Humans. Verified by Agents."

Codex Community Hackathon - Pune

Links

Repository

github.com/neel-develops/PRUDENCE

Website

prudence-vercel-deploy.vercel.app

Demo video

drive.google.com/file/d/14Go1SeIjVsLb3-CQE2SMNl7HxqXHDKbn/view?usp=sharing

Additional info

How was your experience building with Codex?

Building with Codex was a great experience. It helped us move much faster from idea to implementation by assisting with code generation, debugging, architecture planning, and feature development. The platform allowed our team to focus more on solving the problem and refining the product rather than spending excessive time on repetitive coding tasks.

Describe your experience using Loops House as the hackathon platform. What worked well, what challenges (if any) did you face, and what improvements would you like to see?

Loops House provided a smooth and organized experience for managing the hackathon journey. The submission process, team management, and event updates were easy to access and understand. One challenge was that some platform features and instructions could be explained more clearly for first-time participants. Adding more onboarding guidance, real-time notifications, and a dedicated help section would make the experience even better.

Tell us about your overall experience at Codex Community Hackathon Pune.

The overall experience was highly positive and motivating. The hackathon created an environment where participants could learn, build, collaborate, and innovate under time constraints. It was exciting to interact with fellow builders, explore new technologies, and turn an idea into a working solution. The event was well-organized and provided valuable exposure to real-world problem-solving and teamwork.

What could Codex Community improve to create a better experience for participants?

A few improvements that could enhance the participant experience are: More detailed pre-event guidance and documentation. Dedicated mentor office hours for quick technical support. Better visibility of schedules, announcements, and judging criteria. More networking opportunities between participants, mentors, and industry professionals. Additional workshops before the hackathon to help beginners get started quickly.

Team

1 member
  • NE

    Neel Sonawane

    Owner

Overview

🏛️ PRUDENCE - Project Description (No Tables) Executive Summary PRUDENCE is India's first Codex-powered domain agent designed to autonomously verify building plans against NBC 2016 (National Building Code) and RERA regulations.

Mission: Reduce ₹50Cr/year in construction rework costs caused by compliance errors Core capability: Autonomous AI agents read architectural plans and cross-reference thousands of regulation clauses Output: Flags violations with precise citations, risk levels, and fix suggestions Speed: Minutes instead of weeks The Problem Indian architecture and construction firms lose approximately ₹50 Crore annually to compliance-related rework.

Root causes:

Complexity: Over 15,000 pages of NBC 2016 regulations plus state-specific RERA rules Manual Process: Human experts spend 2-4 weeks reviewing a single 50,000 sq. ft. building plan Error-Prone: Manual review misses 15-30% of violations (NITI Aayog 2023 studies) Costly Delays: One compliance error can halt a construction project for 3-6 months Accessibility Gap: Small and mid-sized builders cannot afford expert compliance teams Current solutions fail because:

Manual plan review by architects (hours to weeks) Spreadsheet-based checklist systems (high error rate) Legacy CAD plugins (limited regulation coverage) Government e-approval portals (slow, opaque) The Solution What PRUDENCE actually does:

Ingestion: Builder uploads architectural plan (PDF/CAD formats) plus project type (residential/commercial/industrial) plus location (state) Parsing: Plan Reader Agent extracts layers, dimensions, materials, zoning data using Codex's multimodal reasoning Regulation Matching: Rule Evaluator Agent maps plan attributes to relevant NBC 2016 clauses Violation Detection: Risk Analyzer Agent cross-checks plan against 15,000+ clauses, identifying: Critical: Safety/fire code violations (e.g., missing fire exits) Major: Regulatory breaches (e.g., FSI exceeded) Minor: Technical gaps (e.g., signage not up-to-code) Reporting: Report Generator Agent produces: Executive summary (compliance score: 72/100) Detailed violation list with clause citations Fix recommendations (cost-impact analysis) Risk assessment (project delay, fines, legal exposure)

Technology Stack Agent Platform

Codex (AI coding agent) Enables autonomy, parallel task execution, worktrees Plan Parsing

PyMuPDF plus CAD libraries Extract geometry, layers, metadata Regulation Database

Custom vector index of NBC 2016 plus RERA Fast semantic search of 15K-plus pages Agent Orchestrator

Custom Python plus Codex Skills Coordinate parallel agent workflows Frontend

React plus Tailwind Clean, builder-friendly interface Backend

FastAPI plus PostgreSQL Scalable, reliable API layer ML Models

Custom fine-tuned LLM on construction regulations Domain-specific reasoning

ExploreProjectsMine