Product Walkthrough

A guided overview of the Railway Inventory Intelligence Platform and how this public demonstration is constructed.

Business problem

Railway operators manage thousands of spare-part components across geographically distributed depots. Stockout of even a single critical part can take rolling stock out of service, delay maintenance windows, and force expensive emergency procurement. Most operators still rely on lagging spreadsheets and manual reconciliation across depots and suppliers.

The Railway Inventory Intelligence Platform ingests inventory and procurement documents, normalises them into a relational model, classifies each item by operational risk, and presents a unified decision-support view for procurement, maintenance, and operations teams.

Platform capabilities

  • Document intake with automated extraction & validation
  • Item master + stock + consumption normalisation
  • Criticality classification with configurable thresholds
  • Procurement & purchase-order tracking with delay detection
  • Supplier scorecards & alternate-source pipeline
  • Alert engine with escalation, acknowledgement, and audit
  • Operations copilot for natural-language analysis
  • Email & WhatsApp notification dispatch (production only)

Production workflow

  1. Upload inventory documents (PDF / CSV / XLSX).
  2. Validate schema and data quality before persistence.
  3. Extract inventory, consumption, and procurement records.
  4. Detect critical and near-critical stock against policy thresholds.
  5. Calculate coverage and stockout risk per item & facility.
  6. Identify covered and uncovered procurement requirements.
  7. Generate procurement recommendations and supplier suggestions.
  8. Track purchase orders, inspection milestones, and delivery health.
  9. Surface alerts and exposure analytics for operations leadership.
  10. Support natural-language operational analysis via AI workspace.

Demo architecture

This public demo runs the production Next.js frontend with a sanitized data layer. A single configuration flag (NEXT_PUBLIC_DEMO_MODE) routes every API call to in-browser implementations backed by independently generated fictional records.

The full production backend — FastAPI services, Celery workers, PostgreSQL schema, LLM pipelines, and notification adapters — remains in the repository but is intentionally disconnected from the public runtime.

Synthetic data disclaimer

All inventory items, facilities, suppliers, purchase orders, alerts, documents, and analytics in this demo are independently generated for demonstration purposes. They do not reproduce, paraphrase, or derive from any real operational records.

Identifiers follow obvious fictional namespaces (e.g. INV-RP-1001, SUP-DEMO-008) and supplier names are drawn from a fictional namespace.

Public-demo limitations

  • File uploads and live document parsing are disabled.
  • Authentication & user provisioning use a fixed demo profile.
  • Notifications, WhatsApp, and email integrations are simulated.
  • Database persistence is disabled — temporary actions reset on reload.
  • AI analysis is computed from the local fictional dataset, not a live model.

Workflow video

Workflow video coming soon

The full production workflow video will appear here once available.

Technical architecture overview

FastAPI + SQLAlchemy

Backend services and data persistence

Celery + Redis

Async parsing and notification workers

PostgreSQL

Item master, stock, alerts, audit log

MinIO

Document storage with signed-URL access

OpenAI / Anthropic LLM

Structured extraction & analysis

Next.js 15 + React Query

Operations dashboard frontend