Live Cockpit:sanket.aryanshrivastava.dev ↗

AI-Powered Automatic Block Planning for Indian Railways

Sanket coordinates high-density passenger express traffic with vital infrastructure maintenance windows across the 1,384 km New Delhi – Mumbai corridor. Eliminates operational bottlenecks through in-memory CP-SAT constraint optimization and calibrated failure risk analytics.

Launch Live Cockpit Explore AI Engine Backend & APIs Frontend Docs

Corridor Reach

1,384 km

NDLS to MMCT Golden Quadrilateral

Critical Throughput

+25%

High-priority jobs vs FIFO Baseline

Joint Bundling

40%+

Closure reduction via co-planning

AI Architecture

0ms Latency

In-memory solver inside Django

Corridor Digital Twin & Station Topology

Real-world corridor evidence with automatic alias resolution for historical Indian Railways station codes.

Corridor Digital Twin: New Delhi – Mumbai (1,384 km)

Golden Quadrilateral Route • 10 Critical Junctions with Real Telemetry & Station Aliasing

100% Real Evidence

Tap any station node to inspectSwipe track ➔

[NDLS]

New Delhi

Northern Railway (NR) • Cumulative Distance: 0 km from NDLS

Origin & Traffic Hub

Permissible Speed130 km/h MPS

Track InfrastructureDouble / Quadruple Electrified

System Architecture & Documentation Portals

Select any of the three core sub-systems below to access complete technical guides, schemas, and configurations.

Next.js 16.3 + React 19

Frontend Cockpit (Sanket UI)

Operations control center tailored for Indian Railways section controllers. Features Leaflet GIS mapping, dynamic timetable schedules, multi-category fleet tracking, asset CRUD, and AI recommendation banners with 1-Click approval.

Leaflet GIS map with corridor occupancy
TanStack Query v5 optimistic UI updates
Hydration-safe skeletons & mobile overlay

Frontend Docs Live App

Django 6.1 + DRF

Backend & Intelligence Engine

Unified monolith powering relational persistence (PostgreSQL 16 / SQLite), Celery 5.6+ timetable sync, RailKit quota-guarded train telemetry, and embedded in-memory AI constraint optimization without external microservices.

15+ REST endpoints with IST timezone handling
3-tier auto-overdue task DELAYED lifecycle
35-request Bruno automated test collection

Read Backend & API Catalog

Google OR-Tools CP-SAT

Railway AI & Optimization Engine

Mathematical co-pilot for controllers. Uses Calibrated XGBoost and Cox Survival analysis to score failure risk, and OR-Tools CP-SAT discrete optimization to pack conflict-free maintenance windows across 6h daily, 7d weekly, and 30d monthly horizons.

Multi-department joint bundling (40% fewer closures)
Dynamic operational pressure scoring via train delay
Explainable reason tags (`HIGH_FAILURE_RISK`, etc.)

Read Railway AI Docs

5-Minute Quickstart: Running Locally

Follow these three simple steps to start the complete platform on your workstation.

Launch Unified Backend (Django + Embedded AI)

Terminal: Backend Setup

# Navigate to backend directory
cd backend

# Option A: Fastest with uv
uv pip install -r requirements.txt
uv run python manage.py migrate
uv run python manage.py runserver 8000

# Backend will serve DRF Browsable API at: http://127.0.0.1:8000/railways/

Launch Frontend Cockpit (Next.js 16)

Terminal: Frontend Setup

# In another terminal window
cd frontend

# Install dependencies and start development server
bun install   # or: npm install
bun dev       # or: npm run dev

# Open cockpit in browser: http://localhost:3000

Verify Embedded AI & Test Suite

Terminal: AI Health Check

# Run in-memory CP-SAT & Calibrated XGBoost unit tests
python -m unittest tests/test_embedded_ai.py

# Check engine health in 1 line
python -c "from src.services.ml_engine import RailwayMLEngine; print(RailwayMLEngine().health())"
# Output: Engine status: HEALTHY (Models loaded: 6/6)

Empirical Proof: AI Optimizer vs Traditional FIFO

Validated against Indian Railways corridor traffic simulations across 75 automated test suites.

Verified Benchmark

Evaluation Metric	FIFO Baseline	Railway AI (CP-SAT)	Net Impact
Critical Track Defect Throughput	Baseline (unprioritized)	+25% high-risk tasks scheduled	Prevents rail fractures by resolving urgent defects before deadlines.
Multi-Department Bundling	Near 0% (Siloed department blocks)	40%+ joint track closures	"One Closure, Three Jobs Done": Civil, S&T, and Traction share windows.
Ripple Passenger Delays	High (Random peak closures)	Steered to delay-trough intervals	Dynamically respects Vande Bharat & Rajdhani priority headways.
Decision Explainability	Manual intuition / Paper logs	Plain-English reason tags + Confidence	Human controllers retain 100% oversight and 1-Click approval rights.