Smart City IoT Platform · Marcelo Santamaria

Architecting an edge-computing IoT platform to monitor and manage 50,000+ connected devices across a smart-city deployment.

Constraints

Devices distributed across many physical locations
Intermittent connectivity at the edge
Strict security requirements for public infrastructure
Small architecture team

Problem

City infrastructure — traffic signals, waste-management systems, environmental sensors — needed real-time monitoring and predictive maintenance, but the devices were physically distributed, connectivity was inconsistent, and no central system existed to manage them as a fleet rather than as individual installations.

Constraints

Devices were spread across many physical locations with intermittent connectivity, ruling out designs that assumed constant uptime.
Public infrastructure meant security requirements — encryption, authentication, secure boot — were non-negotiable from day one.
The architecture team was two engineers, so the device-management layer had to be centralized and largely self-service rather than operated by hand.

Decision

The platform was built around edge computing so that devices could make local decisions and buffer data during connectivity gaps, combined with a centralized device-management layer for fleet-wide visibility and control.

flowchart LR
    Sensors[Field Sensors] --> Edge[Edge Gateway]
    Edge --> Cloud[Cloud Platform]
    Cloud --> DeviceMgmt[Device Management]
    Cloud --> ML[Predictive Maintenance ML]
    DeviceMgmt --> Sensors

Security was designed in rather than added later: every device authenticated with unique credentials, all traffic was encrypted end-to-end, and secure boot prevented tampered firmware from joining the fleet.

Trade-offs

Centralizing device management added upfront complexity compared to treating each site as a one-off deployment, but it was the only way two engineers could operate 50,000+ devices without the fleet becoming unmanageable as it grew.

Result

50,000+ IoT devices brought under centralized management across multiple locations.
Improved device uptime and operational efficiency from proactive, fleet-wide visibility instead of reactive, per-site troubleshooting.
A reusable IoT security framework — encryption, device authentication, secure boot — that carried forward into later projects, including the telemetry ingestion pipeline built on top of this platform.

Lessons Learned

Designing for intermittent connectivity from the start is far cheaper than retrofitting it later. Edge computing wasn’t just a performance optimization here — it was what made the platform operable at all, given the realities of public infrastructure networking.