Deploying a UNDP-Funded Emergency Alerting Platform Across the Eastern Caribbean

Background

While at BlackBerry AtHoc (then Alerting Solutions), I worked on a UNDP-funded emergency alerting project covering Anguilla, St. Martin, Montserrat, and Aruba. The requirements went beyond a typical single-site deployment: Common Alerting Protocol (CAP) support, multilingual and international-font support, a mobile-usable interface, secure locally hosted servers on each island, and a support model that had to work remotely once the systems were live.

Approach

I adapted a PHP frontend framework I'd previously built into a mobile-capable interface for the CAPServer alerting platform, so field responders and officials could work with alerts from a phone rather than only a desktop console. Each island ran its own Dell 2U physical server, hosting the platform under VirtualBox managed from the command line, administered remotely over SSH.

Before going live, we validated the system end to end using historical CAP alerts, run all the way through to siren activation, rather than stopping at message delivery.

Deployment

Installation on each island had to be done by local workers, not by a traveling engineer doing every site personally. I traveled to Anguilla and, once there, worked directly with a local technical consultant — walking them through the documented build and configuration process rather than performing the install myself. That consultant then went on to install the systems on the remaining islands, using the same documentation.

The St. Martin Incident

Some time after go-live, remote access and siren activation on the St. Martin system stopped working. Diagnosing it remotely took months without resolution, so I traveled onsite. Once there, finding the cause took about five minutes: a newly installed customer firewall was blocking both SSH and the outgoing siren-activation traffic the system depended on. Fixing it meant working with the customer's local IT to restore the network configuration the deployment had originally documented.

Lessons

The gap between "works when deployed" and "still works months later" is often infrastructure, not application code — a firewall change made for unrelated reasons broke a life-safety system that had nothing to do with it. Documentation and coordination with local IT aren't a formality around the real engineering work; for a distributed, remotely supported system, they're part of what keeps it reliable.

Knowledge transfer mattered just as much as the technology. The Anguilla approach — document the process well enough that a local consultant could execute it, then let them replicate it — scaled the deployment across islands in a way that flying to each one myself never could have.