Project Overview
Supu Technologies engaged FoogleTech Software to design and build a fully on-premise IoT data collection and analysis platform for their distributed field operations. The system connects STM32-based sensor units to Toradex Verdin AM62 SBC devices across 20,000 field locations, aggregates real-time sensor data alongside secondary operations (sales, installations, and inventory transactions) through an on-premises MQTT broker and backend API, and delivers live insights through a custom React analytics dashboard — all without a single byte leaving their private infrastructure.
The Engineering Challenge
- The client operated 20,000+ field-deployed SBC-IoT units and 800 secondary logistics devices across distributed locations — building a scalable on-premise architecture that handled this volume without cloud infrastructure was architecturally complex and operationally critical.
- Full data sovereignty was non-negotiable: cloud-based IoT platforms (AWS IoT, Azure IoT Hub) were explicitly out of scope. The entire stack — edge, broker, backend, and dashboard — had to run on privately owned Linux servers with no external dependencies.
- Field devices operated in areas with intermittent network connectivity. Any data loss during outages was commercially unacceptable — the system needed reliable edge-level caching with automatic sync recovery on reconnection.
- The platform had to unify two entirely different data streams: real-time sensor telemetry from STM32 units (high-frequency, time-series) and secondary business operations data (sales records, installation logs, inventory transactions) — each with different schemas, latencies, and retention requirements.
- Security across 20,000+ distributed endpoints needed to be robust without relying on cloud-managed certificate services: mutual TLS with signed payloads, JWT-authenticated APIs, and role-based access control had to be implemented entirely on self-hosted infrastructure.
Our Solution
- FoogleTech architected a three-tier on-premise edge-to-server pipeline: STM32 microcontrollers gather raw sensor signals and communicate via UART/USB to Toradex Verdin AM62 SBC boards; the SBCs preprocess data, manage RFID scanning, handle offline caching in local SQLite, and sync to the central server via HTTPS with mutual TLS — keeping each daily payload under 5 KB per device for bandwidth efficiency.
- A self-hosted MQTT broker with Python-based messaging services was deployed on-premises for event-driven communication between edge devices and the backend, with Docker containerisation enabling clean service isolation and horizontal scaling across the secondary device fleet of 800 logistics nodes.
- A dual-database backend was designed using PostgreSQL for structured business operations data and TimescaleDB (PostgreSQL extension) for high-frequency time-series sensor data — enabling efficient querying across both data types without schema conflicts or performance degradation at scale.
- An edge caching and sync recovery engine was built into every SBC unit: during network outages, all sensor readings and transaction records are buffered locally and automatically replayed to the server upon reconnection — with conflict detection and deduplication to ensure data integrity across distributed sync events.
- FoogleTech delivered a full React/Next.js analytics dashboard with Recharts and D3 visualisation, covering vendor and outlet registration, real-time sensor monitoring with configurable alert thresholds, daily collection summaries, transaction reconciliation, and historical reporting — all served through Nginx reverse proxy on the client's on-premise Linux servers.
- The 20-week engagement was structured across 6 milestone phases — from architecture and schema design (M1) through SBC and secondary system development (M2), backend API and analytics engine (M3), frontend dashboard (M4), integration and testing (M5), to on-premises deployment and knowledge transfer (M6) — with signed deliverable acceptance gates between each phase.
Results & Outcomes
- End-to-end IoT platform deployed and operational across 20,000 SBC-IoT field units and 800 secondary logistics devices — all data collected, processed, and stored entirely on-premises with no cloud dependency.
- Zero data loss during network outage scenarios: edge caching with automatic sync recovery maintained 100% data integrity across simulated and real connectivity interruptions during acceptance testing.
- Complete data sovereignty achieved: all sensor telemetry, transaction records, and analytics remain within the client's private Linux server infrastructure — fully compliant with their data governance requirements.
- Unified analytics dashboard delivered covering real-time sensor monitoring, outlet management, inventory tracking, daily collections, and historical reporting — reducing manual reporting effort and enabling data-driven operations for field management teams.
- All 6 milestone phases delivered on schedule across the 20-week engagement, with full API documentation, deployment manuals, and structured knowledge transfer sessions ensuring the client's team can independently operate and extend the platform.
Why Supu Technologies Chose FoogleTech Software
FoogleTech Software is a specialist engineering company with over a decade of expertise in AI, embedded systems, IoT, and full-stack software development — serving product teams and enterprises across Industrial Automation and beyond. Our engineers don't just write code — they understand the domain, the constraints, and the real-world pressures that ship deadlines create. For Supu Technologies, that meant deploying a pre-vetted team with direct experience in STM32, Toradex Verdin AM62, Python, reducing ramp-up time from months to days and delivering production-quality work from the first sprint.
Every FoogleTech engagement starts with a structured discovery phase, follows a disciplined agile delivery model with daily engineering syncs, and ends with complete documentation handover — so your in-house team owns the outcome. No black boxes, no lock-in.