Breaking Into AI: Advice From Young Kiwi AI Engineers

Utilising AI is fast becoming the most important skill for business owners, employees and new graduates, yet universities and traditional education are not keeping up with how fast it is moving. That gap is exactly why Young Kiwis in AI exists. In this session, two young engineers at leading New Zealand companies share how they broke in: Georgia Singleton, an AI engineer at Trade Me, and Brodie Dye, a junior AI integration engineer at Zuru.

Georgia Singleton: AI engineer at Trade Me

Georgia's route into AI was anything but linear. She had been interested in AI since she was a kid and started a Bachelor of Science in Computer Science at the University of Canterbury in 2020, in the middle of COVID. After a college advisor told her she was not smart enough to work in AI, she switched to a Bachelor of Commerce majoring in Information Systems with a maths minor.

The break came through a cloud engineering internship at Trade Me in Christchurch. She had always been curious. She still remembers a friend using early AI-assisted code completion back in 2020 and thinking it was amazing, and at uni she took papers on AI and on real versus fake AI. About a year before this talk, an internal role opened on a new AI engineering team at Trade Me, she applied, and she got it.

On what she knew versus what she learned on the fly, she is blunt: university gave her the soft skills and work ethic, but she learned the technical side on the job, because tertiary has not caught up.

A day in the life is two mandatory office days, otherwise remote, on a 40-hour week. Most of her time now goes on architecting solutions through Claude Code rather than writing code herself, working across developers, product managers, owners and designers. Trade Me runs Agile with daily stand-ups, and gives staff 10% learning time, dedicated hours each week to learn something work-related.

Her favourite hack is a Claude scheduled task built from a custom skill. It runs a diagnostic of her existing Python knowledge, looks at her git history and recent projects to find gaps, then Slack messages her a question at a set time each day and tells her if she got it right. Her point: there is huge opportunity to automate your own learning and make it part of your daily routine.

She was on the team that shipped Trade Me's ChatGPT property app last year, New Zealand's first ChatGPT app, built in an intense and very short window after OpenAI announced ChatGPT apps about a week and a half after she joined the team.

Her honest lesson is that it is impossible to keep up with everything in AI. As a self-described perfectionist she wanted to watch every video and read every article and LinkedIn post, but it is not realistic, and you need time to actually absorb what you learn. Give yourself grace to sit with it. Her biggest surprise was that tools like Claude have shifted her core work from developing to architecting far earlier in her career than used to happen. A junior or intermediate engineer would once still be writing a lot of code, but on her small team she is making real architecture decisions.

Her tips: pick one thing at a time and go deep rather than trying to learn every tool, because the skills transfer and most tools are the same thing in a different font; brainstorm problems you face in everyday life, not just at work, and use them as passion projects to get in; and do not wait for the perfect time or opportunity, just start, because that first step is what builds confidence.

Brodie Dye: junior AI integration engineer at Zuru

Brodie has been in his role at Zuru for about six months, three of them as an intern. He chose the AI integration role over a more traditional data science path after seeing how forward-thinking the team was and the mentorship on offer. Zuru's motto, fire bullets first then cannonballs, shapes how the team works: people across the business submit AI ideas, the team spends about three to five days proving value, kills the ones that do not work early, and hands the winners to the full development team.

His own path ran through learning to code in high school, where he built a dating app before you could code with AI, which was brutal but taught him to own something end to end, then engineering science at uni and a few data internships. After graduating he tried life as a stockbroker, did not enjoy it, but developed a strong interest in AI safety and how increasingly capable AI will merge with society. Teaching himself how models are trained and starting to vibe code led him to Zuru. A key takeaway: he applied without thinking he would get it, and his genuine interest plus a little vibe coding was enough, so never let not knowing everything about AI be a blocker.

The role is roughly 50% building and 50% people. He spends a lot of time vibe coding apps and MCP servers, but just as much time translating technical concepts for non-technical marketing, commercial and product teams, and understanding their pain points. With tight three-to-five-day turnarounds, the team only puts real effort behind ideas that prove valuable. A big part of his job is automating himself: anytime he does something more than three times he asks whether he can automate it, and he tries to skillify everything he does.

His tools include Cursor and Claude for coding and planning, Airtable for structured data that AI can read and write easily, Fly.io for hosting simple sites with some control, and Terraform for spinning up cloud infrastructure fast. He has built an internal skill that turns any submitted AI idea into a project he can start building immediately: the idea, the kickoff meeting transcript, the documentation and his own code-base conventions all get pulled in, so Claude already knows to use Terraform, how to handle secrets, and to log every prompt and mistake.

Among his builds: a copy checker that tests product copy against retailer rules and agent engine optimisation; an ingredients compliance checker that verifies Zuru's beauty products against 34 international regulatory bodies, which Claude in Chrome helped derive by visiting the same sites a human would have spent hours on; an internal AI leaderboard that ranks weekly AI use and nudges people to use more skills and connectors; and his intern project, a knowledge assistant that searches Zuru's SharePoint data using semantic search to answer questions.

His lessons are practical. He once accidentally spent about $4,000 on Databricks, because AI tools make it easy to run something many times without thinking about cost, so build good cost dashboards. He learned that building tools is not enough: his manager kept asking for the numbers, how many people used it and how much value it created, and attributing value to AI work is something no one does perfectly yet but everyone will need to. And he learned that there is a difference between people believing a tool works and being excited enough to use it. Something visual and exciting that is clearly relevant to their work drives adoption far more than a thorough explanation. He also keeps a running list of hundreds of mistakes and learnings, because AI is new to everyone.

His tips for getting in, framed partly around what would impress an interviewer: start by coding something you are genuinely passionate about with Claude Code or Cursor; keep up with AI news, where he rates the AI Daily Brief, Epoch's newsletter, the Dwarkesh podcast and the AI Frontiers newsletter; and think in systems, not code. AI is excellent at maths and coding but weaker on the people side. Anyone who can draw a good systems diagram or flowchart can essentially code now, but not everyone who can code can think about how a solution integrates with people and where the human in the loop belongs.

Key takeaways

• University gives you soft skills and work ethic, but you will learn the technical side of AI on the job. Tertiary has not caught up.
• Do not let not knowing everything be a blocker. Genuine interest and a little hands-on building got both engineers hired.
• Pick one tool and go deep. Most tools are the same thing in a different font, and the skills transfer.
• AI has moved junior engineers from writing code to architecting solutions far earlier than before.
• Think in systems, not code, and watch the costs. Adoption comes from tools that are visual, exciting and clearly relevant.