Infrastructure work doesn't produce screenshots. So here's what these actually were: the problem, what I did, and what changed as a result.
Led the move of a high-traffic AWS-hosted client onto Ubuntu 24.04, Apache 2.4 and MariaDB 10.11 with a ~176 GB database. Header-based ALB routing let the new stack serve real traffic for testing before anyone was cut over, backed by a full dump/restore sync and a Gitolite auto-deploy pipeline.
→ Seconds-level target-group rollback path
Resolved sustained load events of 45–80 on 8–24 core production servers by attributing I/O and CPU with pidstat, iostat and sar rather than guessing. Causes traced to scanner bursts running under PHP-CGI without LVE caps, oversized IMAP maildirs holding Dovecot index locks, and EXT4 metadata corruption.
→ Root-caused, not restarted
A disk-full event corrupted InnoDB and forced a MariaDB rebuild on a DirectAdmin server. Recovered order, tax, attribute and shipping data across multiple stores by working from pre-rebuild data-directory snapshots.
→ Live commerce data recovered, not written off
Investigated and contained active compromises involving rogue systemd services, /etc/ld.so.preload hooks, hidden payload fetchers, webshells and crypto-mining processes. Forensic log analysis, cleanup and post-incident hardening across WordPress and Joomla environments.
→ Contained, cleaned, hardened against repeat
Deployed ModSecurity vendor rulesets (OWASP and Comodo) wired to CSF auto-ban triggers, Fail2Ban watching non-standard SSH ports under firewalld, and KernelCare live patching — plus a firewall rule optimisation pass.
→ Fewer repeat incidents across the estate
Executed cPanel-to-cPanel and cross-provider migrations onto AWS EC2, with post-migration validation sweeps — curl status checks across every domain, then path and permission fixes — before the old host was released.
→ Validated before handover, not after complaints
Troubleshot delivery and authentication failures across Exim, Postfix and Dovecot — SPF, DKIM and DMARC records, spam filtering and blacklist delisting — alongside DNS record fixes, certificate renewal and manual AutoSSL recovery via acme.sh.
→ Inbox placement and SSL restored
Maintained Zabbix, Grafana and Nagios coverage across multi-server fleets, tuning alert thresholds so pages meant something, and running proactive health sweeps to catch disk and load problems before customers did.
→ Problems caught before the ticket arrives
Internship project at Exposys Data Labs: an Arduino-based irrigation system using soil-moisture sensors to regulate water levels automatically.
→ Built end to end during a one-month internship
No projects match that filter.
Happy to walk through the diagnosis on any of these — what the symptoms looked like, how it was attributed, and what the fix actually was.