About
I'm Aaron Zhang, a product manager at TikTok LIVE, based in Sydney. I build AI products for the people who run the platform's campaigns and the creators who go live on it.
I've built campaign products for ten years, across telecom, finance, e-commerce and now live streaming. The industry kept changing; the product didn't. Decide who you're trying to move, reach them, then find out whether the number moved because of you.
Half the work is efficiency, taking hours out of a process. The half I care about is effect, and effect is where the arguments start. That's what I write about.
Where I've worked
-
2024 → now
AI products for campaigns and creatorsI build the AI products for TikTok LIVE's campaign operators and creators. Three agents in production; one lifted participation and revenue under a controlled A/B test.
-
2021 → 2024
Marketing and creative platformsI built the campaign platform 0→1, folding ten separate systems into one, and a creative platform that generated about ten thousand AI product images a month, with a patent on the rendering.
-
2021
Marketing algorithms, financial servicesI worked on card-acquisition campaigns, separating hand-written rules from the models so analysts could tune a strategy without a deploy.
-
2019 → 2021
Call-centre marketing AII built live intent and eligibility scoring for two hundred call-centre agents, and the marketing bot inside the banking app.
-
2017 → 2019
Marketing analytics for a US telecom carrierMy first job. I built Report AI, which cut report generation from three hours to forty-five minutes.
-
2013 → 2017
Automation and controlA second degree in finance alongside it.
On my own time
-
DailyWork
Open-source agent skills I run in a real work week: eval gates, blind A/B verification, metric attribution, cross-model review. Built for Claude Code and Codex CLI.
-
Trend Intelligence Engine
A daily digest that reads about seventy sources and hands back the few things worth acting on. I knew nothing about intelligence mining; I owned the question and the standard for checking answers.
-
No Magic, Only Method
A talk about how far you can let AI climb and what you still have to own at every rung. Published as a deck you can fork.
Next
Coach-style AI products: career coaching, study-abroad advising, adjacent things. Same shape as everything above, a person facing a real decision who needs an answer they can argue with. If you're building here, I'd like to compare notes.
Compare notes →