Portrait of Janice Lin

Janice Lin

QA engineer by title, bug detective by nature.

I break AI products and websites on purpose, so your users don’t find the cracks first. Every audit ends with clear findings and a plain go or no-go.

  • LLM evaluation
  • AI safety testing
  • ISTQB Certified QA

Author of How to Test AI Products

About me

Ten years in QA, and I still get a small thrill when a test fails. It means we found the bug before a customer did.

These days I work on generative AI, which means asking a chatbot a lot of awkward questions. Can it be talked into saying something it shouldn’t? Does it make things up when the answer isn’t in its sources? Does it use its tools the way the team meant it to? I also make the release go or no-go calls on its launches.

My method is simple: measure first, then test the way a motivated attacker or an impatient user would, and report what I find in words a product owner can act on.

Before AI, I tested the TikTok Shop seller platform and Amazon Seller Center. Real users will always find the one button you hoped they’d never press. That’s where I learned my favourite rule: assume someone will do the one thing you didn’t plan for, then write down exactly what happened.

I built QA GoGo because the first draft of a test plan shouldn’t eat a whole day. It does the drafting, and the QA engineer keeps the judgement.

What I audit

AI products

Chatbots, RAG systems and AI agents: prompt injection, data leakage, hallucination, unsafe tool use and compliance risk, with a documented release gate.

Websites

Functionality, accessibility, content accuracy and compliance risk, reported with severity, evidence and steps to reproduce.

How an audit works

A short, predictable process, so you know what you get and when.

  1. Scope

    We agree what is being shipped, the risks you care about, and the deadline.

  2. Test

    I run structured and adversarial tests against the product, with every step recorded.

  3. Report

    You get findings ranked by severity, each with evidence and steps to reproduce.

  4. Recommend

    A clear ship, fix-first or hold call, plus a retest once fixes land.

Experience

  • AmazonQA Engineer, generative AI

    Quality and release decisions for a production LLM assistant.

  • TikTokLead QA Engineer

    Led QA for TikTok Shop’s expansion into seven markets and for its fraud-detection model.

Working together

Tell me what you are shipping and when. I will reply with how I would approach the audit and what you would receive.