ikwart/Mantrack-entreprise-platform

An end-to-end data platform for a fictional Cat equipment dealer in Ghana — real Kafka streaming, a trained ML model for predictive maintenance, dbt, RBAC, and an LLM query agent, all built and stress-tested. # Heavy Equipment Entreprise Data Platform An enterprise data platform simulating **Mantrac Ghana** — the authorized Caterpillar dealer in Ghana — across its real business lines: **equipment sales, parts, service, and rental**, integrated with a **VisionLink-inspired telemetry and predictive maintenance layer**. This is a portfolio project demonstrating Data Engineering, Analytics Engineering, Business Intelligence, Cloud/Streaming Architecture, and Industrial IoT — built to show the kind of system thinking relevant to data roles at heavy-equipment dealers and industrial firms (Mantrac, Komatsu, Epiroc, Cummins, and similar). **This is not a toy dashboard project.** It's a full pipeline: synthetic-but-realistic data generation → streaming telemetry → a real trained ML model → automated business-rule triage → a human-in-the-loop review step → parts demand forecasting → orchestrated daily batch processing → a tested analytics layer. Every piece listed as "done" below was actually run against live infrastructure (a real Postgres database, a real trained Spark ML model, real dbt test runs) during development — not just written and assumed to work. --- ## Table of Contents - What this platform simulates - Architecture - Repository structure - Quick start - Walkthrough: what each layer does - Data model summary - Business context --- ## What this platform simulates Mantrac's real business runs on four revenue legs — equipment sales, parts, service, and rental — across mining, construction, quarrying, and oil & gas/energy customers in Ghana. This platform models that end to end: - **A synthetic but realistic fleet**: 18 real Ghanaian companies (Gold Fields, AngloGold Ashanti, Newmont, Rocksure International, Engineers & Planners, and others across mining/construction/quarrying/logistics), operating 347 pieces of Cat equipment across 39 models and 12 equipment categories. - **Telemetry that actually degrades realistically**: equipment doesn't fail randomly, a logi …