GLORY SOFTWARE · CASH HANDLING

UBIQULAR Inform

Operator telemetry for a fleet of cash-handling machines: MQTT ingestion into Kafka, NestJS consumers, PostgreSQL read models — and a 30–50% cut in API response time.

Role
Senior Software Engineer · Sub-team Lead
Company
Glory Software Vietnam
Period
03/2025 — 03/2026

Problem

Cash-handling machines in the field emit a constant stream of device, inventory and fault telemetry. Operators needed one view of what was happening across the estate, and the query paths behind that view were too slow to sit in front of a dashboard.

Solution

Ingest device events over MQTT into Kafka, process them in workers, and serve the operator view from a read path tuned for it. Redis carries hot state across instances so the distributed deployment stays consistent. MQTT simulation tooling plus JMeter load tests set the performance baseline before release.

Architecture

Tech stack

  • NestJS
  • Kafka
  • MQTT
  • PostgreSQL
  • Redis
  • RxJS
  • JMeter

Impact

30–50%
faster API responses after SQL and query tuning
0
open pentest findings at handover

More work