🇮🇳 ProjectSetu — Integrated Infrastructure Intelligence

AI-Powered Predictive Monitoring, High-Res Satellite Geospatial Radar & Automated Intervention Platform

Smart India Hackathon (SIH26103) • Ministry of Statistics & Programme Implementation (MoSPI) • IPMD

Transforming India's ₹100+ Lakh Crore National Infrastructure Pipeline (NIP) & PM Gati Shakti into a predictive, transparent, and accountable delivery ecosystem.

🌐 Live Web Platform • 🛰️ Satellite Radar View • 🏛️ Citizen Portal • 📖 Documentation • ⚡ Quickstart

---

📌 Executive Summary

India's mega infrastructure investments—spanning highways (NHAI), railways (RVNL), metros, bulk water pipelines, and power evacuation corridors—frequently encounter substantial schedule slippages and cost overruns. Traditional oversight relies on static PDF submissions, disjointed inter-agency communications, outdated milestone baselines, and post-facto audit postmortems.

ProjectSetu reimagines national project governance by implementing a continuous, closed-loop intelligence cycle:

flowchart LR
    A[📥 COLLECT<br/>Observations & Geo-Telemetry] --> B[🔍 VALIDATE<br/>Evidence & CAPEX Cross-Audit]
    B --> C[🧠 PREDICT<br/>Calibrated Dual-Engine ML]
    C --> D[💡 EXPLAIN<br/>SHAP Attribution & Freshness Lens]
    D --> E[🎯 PRIORITIZE<br/>Composite Risk Scoring]
    E --> F[⚡ INTERVENE<br/>SLA-Enforced Escalations]
    F --> G[✅ MEASURE<br/>Decision Replay & Verification]
    G --> A

Loading

---

🌟 Key Innovations & Differentiators

Innovation	Operational Impact	Governance Benefit
🛰️ Exclusive High-Res Satellite Radar	Real-time orbital imagery via Esri Multispectral tiles paired with dark-line infrastructure corridors and anti-overlap smart telemetry pins.	Eliminates reliance on watermarked map services; provides instant ground-truth spatial context.
🛡️ Cryptographic Risk Passport	Generates an immutable, SHA-256 fingerprint snapshot recording input features, model version (v1.2.0-xgb), policy weights, and officer determinations.	Guarantees non-repudiation; allows officers to inspect past reviews without dynamic recomputation.
⏱️ Freshness & Evidence Lens	Decouples data currency (e.g. 45-day reporting lag) from intrinsic structural engineering risks.	Prevents false alarms on physically sound projects while highlighting delinquent reporting agencies.
⏪ Decision Replay Narrative	Reconstructs the 5-step institutional memory: Observation → AI Prediction → Officer Deliberation → Assigned Escalation → Verified Outcome.	Comprehensive institutional accountability for Parliamentary Committees and CAG audits.
🕸️ Dependency Cascade Graph	Graph-traversal engine mapping critical prerequisite chains: Land Acquisition → Forest Clearance → Utility Shifting → Civil Construction → Grid Sync.	Early warning indicator for downstream bottlenecks before execution begins.
🎛️ What-If Scenario Sandbox	Real-time recovery simulation allowing officers to evaluate contractor augmentation, fund injections, and fast-track statutory clearances.	Quantifies expected risk reduction delta before capital allocation.
👥 Cross-Portal Interconnection & Simulation	Live interconnected portals for MoSPI Officers, Line Agencies, CAG Auditors, Field Contributors, and Citizens with real-time sync.	Multi-stakeholder coordination with built-in action simulation workflows.

---

🖥️ Platform Screens & Feature Walkthrough

1. 🛰️ National Satellite Geospatial Radar (/app)

• Orbital Multispectral Terrain: Powered by Esri World Imagery with zero API key dependencies and zero watermarks.
• Dark Infrastructure Corridors: High-contrast obsidian outlines (#020617, weight 8) with dynamic cyan radar cores showing India's critical transit spines (Delhi-Katra Expressway, Western DFC, Eastern DFC, North-South Spine, and Saurashtra Coastal Pipeline).
• Anti-Overlap Smart Pins: Responsive markers adapt smoothly between zoom levels—sleek 28px circular target pins at national zoom to prevent badge collision, expanding into rich telemetry cards with #proj-code, budget in ₹ Cr, and risk rating on hover or selection.
• Right Telemetry Drawer: Instant access to delay forecasts, physical vs. financial progress gaps, contractor rating, and direct drilldown links.

2. 🏛️ MoSPI National Portfolio Command Center (/app)

• Key Metrics Overview: Total Capital Outlay, Overrun Exposure, Milestone Slippages, and National Average Health Index.
• Multi-Dimensional Filtering: Slice portfolio across ministries (MoRTH, Railways, Power, MoHUA, Jal Shakti), states, risk categories, and physical progress bands.
• Dual Perspective: Seamless 1-click toggle between Geospatial Satellite Radar and High-Density Tabular Data Grid.

3. 🔍 Explainable AI & Project Dossier (/app/projects/:id)

• Baseline Revisions: Track original vs. current sanctioned outlay with cost escalation percentages and authorized revision justifications.
• SHAP Feature Attributions: Transparent top-5 delay drivers (e.g., Inter-tidal Marine Piling Constraints, Land Acquisition Delays, Material Cost Inflation).
• Evidence Verification Queue: Cross-audit comparing physical execution against cumulative financial disbursements to detect anomalous billing.
• Printable Dossier: One-click generation of the official MoSPI Project Monitoring Dossier ready for review meetings.

4. ⚡ Inter-Agency Escalation & Intervention Matrix (/app/interventions)

• Multi-Tier Escalation: Rule-based routing to District Magistrates, State Chief Secretaries, Railway Boards, or Cabinet Committees.
• SLA Enforcement: Target resolution dates with automatic overdue warnings and escalation triggers.
• Evidence-Backed Closure: Line agencies must submit geotagged photos or formal statutory clearances before an officer can mark an intervention verified.

5. 🏗️ Contractor 360 Evaluation (/app/contractors)

• Performance Rating Index: Composite rating (1.0 to 5.0) evaluated across on-time delivery, cost adherence, safety compliance, and dispute history.
• Active Exposure Analysis: Real-time visibility into contractor workload across multiple simultaneous national projects to detect bandwidth strain.

6. 🌐 Public Transparency & Citizen Grievance Portal (/public)

• Open Governance: Unauthenticated public access to whitelisted project summaries, physical progress meters, and geotagged project locations.
• Citizen Feedback & Grievance Lodging: Citizens can submit geo-localized observations or concerns directly to MoSPI with automated tracking tokens.

---

🏗️ System Architecture

graph TD
    subgraph Client Layer [Modern Web Frontend - React 18 + TypeScript + Vite]
        UI[Landing Page & Portal Switcher]
        MAP[Esri Satellite Radar + Leaflet]
        CMD[Officer Command Center]
        AGENCY[Agency Submission Workspace]
        AUDIT[CAG Audit Trail & Dossier]
        PUB[Citizen Transparency Portal]
    end

    subgraph Intelligence Engine [Predictive ML & Decision Logic]
        ML[Python FastAPI Microservice v1.2.0-xgb]
        XGB_REG[Gradient Boosted Delay Regressor]
        XGB_CLS[Calibrated Cost Overrun Classifier]
        SHAP_EXP[SHAP TreeExplainer Attributions]
        FALLBACK[Deterministic Rule-Based Engine]
    end

    subgraph Data & Storage [Persistence & Integrity]
        STORAGE[Local Persistent Storage / Firebase Backend]
        RBAC[Fine-Grained Role-Based Access Control]
        AUDIT_LOG[SHA-256 Cryptographic Audit Ledger]
    end

    UI --> CMD & MAP & AGENCY & AUDIT & PUB
    CMD --> ML
    ML --> XGB_REG & XGB_CLS & SHAP_EXP
    ML -. Fallback Mode .-> FALLBACK
    CMD & AGENCY & AUDIT --> STORAGE
    STORAGE --> RBAC & AUDIT_LOG

Loading

---

🧠 Machine Learning Engine (v1.2.0-xgb)

1. Leakage-Free Temporal Group Strategy

• Synthetic Training Corpus: 200 diverse infrastructure projects across 5 sectors (Highways, Rail, Metro, Energy, Water) generating 4,741 longitudinal monthly observations.
• Strict Group Splitting: Grouped by project_id (80% Train, 20% Test) to ensure no temporal leakage occurs between training features and test outcomes.

2. Dual Predictive Targets

1. Cost Overrun Classifier: Calibrated probability $C \in [0, 1]$ of completed expenditure exceeding the baseline by $\ge 10%$.
   • ROC-AUC: 0.8037 • PR-AUC: 0.8172 • Brier Score: 0.1420
2. Schedule Delay Regressor: Predicts positive slippage in days $S \in [0, \infty)$ beyond the contractual completion date.
   • MAE: 158.8 days • Median AE: 122.5 days • 80% Prediction Interval

3. MoSPI Composite Risk Score Formula

$$\text{RiskScore} = \text{round}\Big(100 \times \big(0.40,C + 0.30,S_{\text{norm}} + 0.20,M + 0.10,B\big)\Big)$$

Where:

• $C$ = Calibrated Cost Overrun Probability
• $S_{\text{norm}} = \text{clamp}\left(\frac{\text{Predicted Delay Days}}{\max(90,, 0.25 \times \text{Original Duration Days})},, 0,, 1\right)$
• $M$ = Milestone Slippage Ratio
• $B$ = Review Blocker Severity Index

---

💻 Tech Stack & Languages

Layer	Technologies
Frontend Framework	React 18, TypeScript 5.3, Vite 6 (Blazing Fast HMR & Tree-Shaking)
Styling & Motion	Tailwind CSS 3.4, Lucide Icons, GSAP Motion, Framer Motion
Geospatial & Mapping	Leaflet 1.9, Esri World Imagery, ArcGIS REST Services, SVG Vector Overlays
State & Data Store	React Hooks, Modular Context, IndexedDB / LocalStorage, Firebase Cloud Firestore
Predictive Backend	Python 3.11, FastAPI, scikit-learn, XGBoost, SHAP, Uvicorn
DevOps & Hosting	Vercel Edge Network, GitHub Actions CI/CD, Docker, Monorepo Workspaces

---

⚡ Quickstart & Local Development

Prerequisites

• Node.js v18+ or v20+ / npm v10+
• Python 3.10+ (Optional, for training/running ML microservice)
• Git

1. Clone the Repository

git clone https://github.com/codexanjan/projectsetu.git
cd projectsetu

2. Install Monorepo Dependencies

npm install

3. Launch the Web Platform

cd apps/web
npm run dev

Open http://localhost:5173 in your browser.

4. (Optional) Run the Python ML Microservice

cd services/ml
python -m venv venv
# Windows:
venv\Scripts\activate
# macOS/Linux:
source venv/bin/activate

pip install -r requirements.txt
python -m uvicorn app.main:app --port 8000 --reload

Interactive API Docs available at: http://localhost:8000/docs

5. Build for Production

cd apps/web
npm run build

---

🎯 Smart India Hackathon Demonstration Walkthrough

1. Landing Page (/): View the mission statement, live satellite radar preview, and portal selection cards.
2. Interactive Role Switcher (/app): Instant 1-click switching between MoSPI Officer, Agency Manager, CAG Auditor, and Field Contributor.
3. Satellite Radar (/app): Explore India's 5 major multi-modal transit corridors and click pins to inspect live telemetry.
4. Project Deep Dive (/app/projects/proj-nhai-001): Inspect the Risk Passport, Baseline Comparison, SHAP Drivers, and Scenario Workspace.
5. Contractor Analysis (/app/contractors): Review performance scores, ongoing commitments, and delivery speed across agencies.
6. Citizen Portal (/public): Test public-facing transparency with unauthenticated access.

---

🏷️ Keywords & Hashtags

#SmartIndiaHackathon #SIH2024 #SIH26103 #MoSPI #PMGatiShakti #NationalInfrastructurePipeline #GeospatialAI #SatelliteImagery #MachineLearning #ExplainableAI #XGBoost #ReactJS #TypeScript #Vite #TailwindCSS #Leaflet #PublicProcurement #InfrastructureMonitoring #GovTech #DigitalIndia

---

📄 License & Ethical Disclosures

This project is open-source software licensed under the Apache License 2.0.

Statutory Disclaimer: ProjectSetu is an AI-assisted decision-support prototype engineered for the Smart India Hackathon (SIH26103). Project records, financial figures, contractor evaluations, and risk classifications represented in demonstration datasets are synthetic and designed for educational and analytical purposes.

---

Built with ❤️ for Digital India & Smart India Hackathon 2024
_{Ministry of Statistics and Programme Implementation (MoSPI) • Infrastructure and Project Monitoring Division (IPMD)}