ConstructIQ is a next-generation predictive intelligence platform that unifies BIM geometric validation, cost overrun forecasting, progress S-curve tracking, and site energy optimization into a single glassmorphic command center — narrated in natural language by Google Gemini.
In the $13 Trillion global construction sector:
- 80% of projects overrun their allocated budgets.
- 70% experience severe milestone delays.
- Project managers juggle 5+ disjointed tools (Navisworks for BIM, Primavera/MS Project for schedules, ERPs for costs, and SCADA/BMS for site power).
- Risks are discovered reactively after physical damage or delay has already occurred.
ConstructIQ fuses telemetry and machine learning predictions across 4 critical pillars:
- BIM Geometric Clash Validator: Runs Axis-Aligned Bounding Box (AABB) spatial overlap algorithms combined with Random Forest + LSTM edit history fault scoring and IsolationForest spatial anomaly detection.
- Cost Overrun Predictor: Uses an XGBoost regressor on static project attributes stacked with an LSTM sequence network on weekly cash flows to output an calibrated probability and percentage drift.
- Progress & Schedule AI: Uses a hybrid CNN-LSTM network to model real-world S-curves, identify critical-path slippage in days, and flag milestone risks.
- Site Energy & IoT Telemetry: Ingests heavy equipment and sensor telemetry to predict power loads with LSTM, detects rogue standby spikes with an IsolationForest, and computes automated greedy shutdown schedules.
- 🧠 Killer Feature — The AI Risk Narrator: An LLM reasoning layer (powered by Google Gemini 1.5 Flash) that ingests structured outputs from all 4 models, diagnoses root causes, and delivers an executive briefing with prioritized action items.
CONSTRUCTIQ ARCHITECTURE
┌──────────────────────────────────────────────────────────────────────────────────┐
│ React + Vite Command Center │
│ • Unified Risk Score Gauge (0-100) • AI Risk Narrator (Streamed Brief) │
│ • Progress S-Curve (Planned vs Actual) • Cost Overrun Forecast Curve │
│ • BIM AABB Geometric Clash Matrix • Site Energy & IoT Telemetry Panel │
│ • ML Flagged Model Components • Real-time AI Copilot Chat Drawer │
└────────────────────────────────────────┬─────────────────────────────────────────┘
│ REST / SSE Stream (/api/analyze/all)
▼
┌──────────────────────────────────────────────────────────────────────────────────┐
│ FastAPI Backend Router │
│ /api/health /api/projects /api/analyze/* /api/chat /api/stream-narrative │
└────────┬───────────────────┬───────────────────┬───────────────────┬─────────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ BIM │ │ Cost │ │ Progress │ │ Energy │
│ Validator │ │ Predictor │ │ Tracker │ │ Optimizer │
│ • AABB 3D │ │ • XGBoost │ │ • CNN-LSTM │ │ • LSTM Load │
│ • RandForest│ │ • LSTM Seq │ │ • S-Curve │ │ • IsoForest │
│ • IsoForest │ │ • Stacking │ │ • Delay Prob│ │ • Greedy Opt│
└──────┬───────┘ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘
└───────────────────┴───────────────────┴───────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────────────────────────┐
│ Cross-Domain Risk Scorer & AI Narrator │
│ • Composite Risk Index (Cost 35% + Progress 30% + BIM 25% + Energy 10%) │
│ • Google Gemini 1.5 Flash Executive Briefing & Interactive Q&A Assistant │
└──────────────────────────────────────────────────────────────────────────────────┘
├── backend/
│ ├── modules/
│ │ ├── bim_validator.py # AABB geometric clash detector + RF/LSTM/ISO models
│ │ ├── cost_predictor.py # XGBoost + LSTM stacking ensemble cost forecast
│ │ ├── progress_tracker.py # CNN-LSTM hybrid progress S-curve forecaster
│ │ └── energy_optimizer.py # LSTM load predictor + IsolationForest shutdown opt
│ ├── services/
│ │ ├── risk_scorer.py # Unified 4-pillar composite risk calculator
│ │ └── ai_narrator.py # Gemini 1.5 LLM narrator & interactive project Q&A
│ ├── main.py # FastAPI app, CORS, routes & SSE stream
│ └── requirements.txt # Python dependencies
├── frontend/
│ ├── src/
│ │ ├── components/
│ │ │ ├── landing/ # Awwwards-grade 3D WebGL landing experience
│ │ │ │ ├── HeroScene.jsx # Three.js 3D digital twin (crane, site mesh, floor levels)
│ │ │ │ ├── HeroHUD.jsx # Technical holographic HUD overlay & live telemetry
│ │ │ │ ├── LandingNav.jsx # Glassmorphic navigation with system health pill
│ │ │ │ ├── IntelligencePillars.jsx # 4-pillar deep dive with interactive cards & live metrics
│ │ │ │ ├── AIRiskNarratorSection.jsx # Simulated streaming Gemini reasoning terminal
│ │ │ │ ├── DashboardPreview.jsx # Isometric 3D command center perspective preview
│ │ │ │ ├── RiskEngineSection.jsx # Mathematical composite risk formula breakdown
│ │ │ │ ├── ArchitectureFlow.jsx # End-to-end topological pipeline visualization
│ │ │ │ ├── TechStackSection.jsx # High-performance tech stack showcase
│ │ │ │ ├── FinalCTA.jsx # Cinematic call-to-action section
│ │ │ │ └── LandingFooter.jsx # Industrial-grade telemetry footer
│ │ │ ├── AIBrief.jsx # AI Narrator with streaming typewriter & copy
│ │ │ ├── AIChatModal.jsx # Interactive AI copilot chat assistant
│ │ │ ├── ClashTable.jsx # BIM clashes matrix with floor/severity filters
│ │ │ ├── CostChart.jsx # Budget baseline vs projected overrun curve
│ │ │ ├── EnergyPanel.jsx # Site consumption, anomaly alerts & shutdown savings
│ │ │ ├── FlaggedTable.jsx # Multi-model component fault ranking with action tiers
│ │ │ ├── Icons.jsx # Crisp SVG icon components
│ │ │ ├── ProgressChart.jsx # Planned vs Actual vs Forecasted S-curve
│ │ │ └── RiskGauge.jsx # Dynamic SVG multi-pillar risk gauge
│ │ ├── pages/
│ │ │ ├── Dashboard.jsx # Command center view with live analytical telemetry
│ │ │ └── Landing.jsx # Awwwards-tier landing page with GSAP entrance & 3D twin
│ │ ├── App.jsx # View router
│ │ ├── index.css # Ultra-premium glassmorphic dark design system & HUD tokens
│ │ └── main.jsx # React DOM mount
│ ├── index.html # SEO, Google Fonts & viewport metadata
│ ├── vite.config.js # Vite configuration & backend proxy
│ └── package.json # Node packages (Three.js, GSAP, React)
├── BIMmodelvalidator.py # Original script (preserved)
├── constructionprogress.py # Original script (preserved)
├── costoverrun.py # Original script (preserved)
├── lstm.py # Original script (preserved)
└── README.md # Platform documentation
# Navigate to backend directory
cd backend
# Install Python requirements
pip install -r requirements.txt
# (Optional) Provide your Gemini API key for live LLM generation:
# set GEMINI_API_KEY="your-api-key-here"
# Start FastAPI server
uvicorn main:app --reload --host 127.0.0.1 --port 8000Backend API will be accessible at: http://127.0.0.1:8000
Interactive Swagger docs: http://127.0.0.1:8000/docs
# Navigate to frontend directory
cd frontend
# Install dependencies
npm install
# Start Vite dev server
npm run devOpen your browser at: http://localhost:5173
- 100% Codebase Reuse: Retained and upgraded all 4 original machine learning scripts into clean, modular, callable backend pipelines.
- Ensemble Depth: Combines Random Forests, XGBoost, Deep LSTMs, Hybrid CNN-LSTMs, and Isolation Forests.
- Intelligent Synthesis: Doesn't just dump charts — the Gemini AI Risk Narrator connects how BIM clashes cause schedule delays, driving up subcontract costs.
- Enterprise-Grade UI: Custom dark-mode glassmorphic interface with interactive SVG gauges, S-curves, and real-time AI assistant.