Hardening a 566,000-user platform with zero downtime.
A high-volume online training platform, running on WordPress and LearnDash with a userbase past 566,000 records, was in a fragile state. Repeated crashes and critical errors, slow admin screens, certificate-generation problems, heavy spam registration, and real security exposure. The user table alone had swollen to a size that made even routine administration risky.
The brief was not to patch isolated bugs. It was to stabilise a live platform that hundreds of people were actively using, without taking training, certificates or access down in the process.
I treated it as a live-service hardening exercise, not a maintenance ticket. I repaired and stabilised the platform, upgraded the runtime and fixed the certificate and PDF issues that surfaced from it, moved scheduled jobs to a proper server-side schedule, and cut heavy admin-side components that were dragging performance.
The standout decision was the userbase. Rather than a destructive cleanup, I built a reversible archive-and-quarantine process, with exports and batch processing, so database pressure could be cut with no risk to genuine learner data. A full security scan surfaced issues for triage, weak passwords were reset, an abused public registration route was closed, exposed backups were moved out of reach, and public access to sensitive server ports was shut, all while keeping the site live.
On top of that I wrote custom admin tooling inside WordPress for managing completions, expiries and certificate downloads, and for controlled review of spam records at scale.
The platform moved from fragile, overloaded and abuse-prone to controlled and supportable, and every bit of the hardening was applied without taking the public site offline. Post-hardening checks still returned a healthy homepage, login and admin.
Registration abuse was contained, the risky services were shut, administration became safer, and the client gained proper structured tooling for managing learner records and certificates at scale. The work also laid a safe, reversible path for the next phase, rather than forcing risky changes all at once.
Risk judgement on a live system. The senior instinct to make everything reversible, quarantine rather than delete, harden without downtime, so a platform with more than half a million users got materially safer without a single moment of broken service for its learners.
That "will not break your live system" credibility is worth more than any single feature.