Work
How I got here.
How a second-year student ended up building software that a working firm depends on, and the people who can vouch for it. The route in was to be useful before being paid.
In their words
Three people who have watched the work up close, asked to write whatever they think. Nothing goes in these spaces until they have written it and approved it.
How this started
Nobody hands a second-year student client files. The way in was to be useful before being paid.
Early 2025
Worked at Harkness AI for nothing
Turned up at a Christchurch AI studio with no commercial experience and offered to work for free, because the alternative was practising on problems nobody had. Unpaid buys you something a course cannot: a real client, a real deadline, and somebody senior who will tell you when the thing you built is wrong.
May 2025
It turned into a part-time job
Junior AI Engineer, part-time alongside the engineering degree. Automation work for a dozen small New Zealand businesses, taken from brief to deployment on my own: invoices captured and pushed into Xero, an accounting system fed by a scheduled pipeline, an inbox classified and written into a knowledge base.
Through 2025
Learned which automation survives
The pattern across all of it was the same. The systems still running a year later are unglamorous and specific. The clever general agents are the ones quietly switched off. That is the single most useful thing the job taught me, and it is why nothing on this page is a platform.
2026
Started taking my own work
Three tools now in weekly use at a Christchurch engineering consultancy, built as Taylor Consulting rather than as an employee. Documents with professional liability attached, which behave very differently from an internal automation: somebody signs them, so a plausible-looking output is worse than no output.