This RFID platform supports warehouse traceability, asset management, and industrial automation. It integrates RFID label printers, multiple reader types, edge devices, and cloud services for real-time EPC visibility, print automation, and operational reporting.

I joined after the proof of concept and worked in a small team with one hardware engineer and outsourced assembly staff. Beyond circuit and production management, I was primarily responsible for device firmware, applications, cloud services, and customer deployment, taking the system into real production environments.

Relational data model and GS1 EPC

The prototype relied heavily on NoSQL and ad hoc schemas. As EPC scans, print jobs, and asset records grew, query performance and consistency became bottlenecks. I led a migration to relational models built with Prisma and MySQL, redesigned import and synchronization across customers, sites, devices, labels, and print jobs, removed duplicate or corrupted records, and standardized state transitions.

I later replaced random EPCs with GS1 SGTIN-96 encoding, defined a compatible EPC prefix, and generated labels from customer asset data. This created precise mappings among labels, scans, and assets for traceability and reporting.

Modular RFID hardware

The prototype combined a Raspberry Pi 4, breadboards, and reader modules. I planned modular compute, power, battery, and extension boards so readers, antennas, batteries, and sensors could be replaced or expanded for different warehouse and food-service environments.

I handled requirements, customer communication, module evaluation, the initial schematic, and hardware/software integration validation, and supported third-party EMC testing and FCC/CE certification preparation.

Embedded Linux firmware and device management

The initial firmware prioritized demonstrations. I migrated the 2019 32-bit Raspbian base to 64-bit Raspberry Pi OS, built a Chromium kiosk with automatic startup, remote maintenance, monitoring, and updates, and preserved compatibility across old and new hardware and 32/64-bit releases.

The system supported multiple printer protocols and RFID readers. Workflow optimization reduced printing to approximately two seconds per RFID label while supporting multi-printer queues. EPC scans were buffered in in-memory SQLite and summarized on demand; Redis Pub/Sub separated drivers from applications so application load would not compromise timely scan capture.

A web UI exposed antenna connectivity, EPC scans, print jobs, LTE RSSI, temperature, battery state, and system events. Devices uploaded normalized events, sensor data, and EPC reports through a Unified DataHub with offline buffering, retransmission, graceful shutdown, battery deep sleep, and USB export—preserving data through unreliable networks and frequent warehouse power loss.

DataHub and production operations

I contributed to the DataHub, worker services, Redis jobs, and cloud data flows. I resolved duplicate data and queue issues, MySQL storage exhaustion, out-of-memory conditions, token expiration, and network incidents, and built backup, monitoring, cron, and anomaly-detection workflows.

For ERP integration, I connected the DataHub to Kafka event streams and converted among multiple schemas for asset tracking, inventory, warehouse automation, and reporting. Production and development deployments were separated to protect customer operations from pre-shipment testing.

I maintained the full print-job lifecycle—from bulk creation, assignment, validation, and preview through shipment and export—and added high-volume CSV import, deduplication, filtering, GS1 decoding, asset matching, and reconciliation to reduce manual warehouse work.

Most customer deployments were in the United States. From Taiwan, I remotely supported rollout, firmware updates, synchronization, and incident analysis. Key EPC, print-job, sensor, and system events were also forwarded to the team’s collaboration channel so deployment staff could coordinate investigation in real time.