Retrieval-grounded support assistant

The problem

Describe the support load before: ticket volume, repetitive questions, response times.

Approach

Structure-aware chunking so answers cited whole sections rather than sentences severed mid-thought.
Hybrid retrieval, keyword and vector combined, because pure vector search performed poorly on exact product codes.
An evaluation set built from real historical tickets, scored automatically, so every prompt change could be shown to help rather than assumed to.
Automatic escalation on low retrieval confidence, on repeat failure, and on any topic flagged as sensitive, with the transcript passed to the agent.
Per-user rate limits and a hard daily spend cap enforced in code.

More work

Backend & infrastructure

Multi-tenant Laravel platform

A Laravel application serving multiple client organisations from one codebase, with per-tenant data isolation and queued background processing.

LaravelMySQLRedisHorizon +2

Server engineering

cPanel hosting fleet migration

Moving a set of cPanel accounts onto new infrastructure, websites, databases, mailboxes and DNS, with the cutover planned so downtime stayed in minutes.

WHMcPanelIMAPSyncrsync +2

Performance engineering

E-commerce performance rescue

A WooCommerce store where the caching plugin had been maxed out and the site was still slow, diagnosed to the actual bottleneck rather than the assumed one.

WooCommerceMySQLRedis object cacheCloudflare +1