Eleven AI roles. What each one
actually does all day.
十一個 AI 職位。
每一個，實際上每天在做什麼。
十一个 AI 职位。
每一个，实际上每天在做什么。
Written for business owners, not recruiters. Each profile tells you the real daily work, the skills that separate a good hire from a confident interview, the salary band, and the certification we require. Click any role to see the full profile including how we test it. 為企業老闆而寫，不是為招聘人員而寫。每個檔案說明真實的日常工作、哪些技能能分辨出「好的人才」與「只是面試講得好」、薪酬區間，以及我們要求的認證。點擊任何職位可看完整檔案，包括我們如何測試。 为企业老板而写，不是为招聘人员而写。每个档案说明真实的日常工作、哪些技能能分辨出「好的人才」与「只是面试讲得好」、薪酬区间，以及我们要求的认证。点击任何职位可看完整档案，包括我们如何测试。
Machine Learning Engineer
Builds, trains and ships the models that power products — from training pipeline to production inference.
機器學習工程師 · 机器学习工程师
LLM / GenAI Engineer
Turns large language models into working products: RAG systems, agents, evaluation and guardrails.
大模型應用工程師 · 大模型应用工程师
MLOps / AI Platform Engineer
Keeps AI systems running: deployment, monitoring, scaling, cost and governance infrastructure.
MLOps / AI 平台工程師 · MLOps / AI 平台工程师
Data Scientist
Turns business questions into measurable models and evidence — forecasting, segmentation, causal analysis.
數據科學家 · 数据科学家
Data / AI Data Engineer
Builds the data foundation AI depends on — pipelines, quality, labelling and governance.
數據工程師（AI 資料） · 数据工程师（AI 数据）
AI Product Manager
Decides what to build with AI and, equally important, what not to build — owns scope, evaluation and risk.
AI 產品經理 · AI 产品经理
AI Governance & Risk Lead
Makes AI defensible: policy, model inventory, impact assessment, audit readiness for regulators and clients.
AI 治理與風險主管 · AI 治理与风险主管
AI Solutions Architect
Sits between the client and engineering: scopes the solution, proves feasibility, designs the integration.
AI 解決方案架構師 · AI 解决方案架构师
AI Enablement / Adoption Specialist
Gets the rest of the company actually using AI — training, workflow redesign, adoption measurement.
AI 導入與賦能專員 · AI 导入与赋能专员
AI Research Scientist
Advances the method itself — new architectures, training techniques, evaluation science.
AI 研究科學家 · AI 研究科学家
AI Trainer / Instructor
Part-time, paid by the hour. Teaches people to operate general-purpose AI agents — WorkBuddy, Qwen, Manus and similar tools — in structured courses with a real assessment at the end.
AI 培訓師／講師 · AI 培训师／讲师
What these roles actually cost 這些職位的真實成本 这些职位的真实成本
Monthly base salary in HKD, excluding bonus and equity. Based on mandates we have run and market data collected in the last twelve months. Use it to sanity-check your own budget before you open the search. The AI trainer line is the exception: that role is part-time and paid by the hour, so its figure is HKD per hour and is not comparable with the monthly bands. 基本薪酬，港幣計，不含獎金與股權。根據我們近十二個月處理的委託及市場數據整理。用來在開始招聘前，先檢視你自己的預算是否合理。AI 培訓師一行為例外：該職位屬兼職、按小時計酬，故數字為每小時港幣，不能與其餘月薪區間直接比較。 基本薪酬，港币计，不含奖金与股权。根据我们近十二个月处理的委托及市场数据整理。用来在开始招聘前，先检视你自己的预算是否合理。AI 培训师一行为例外：该职位属兼职、按小时计酬，故数字为每小时港币，不能与其余月薪区间直接比较。
HKD / hour
Bars are proportional to the range within each role; they are not positioned on a shared axis. Ranges widen with seniority — the top of each band assumes 8+ years of directly relevant experience. 長條長度代表該職位的薪酬跨度，並非在共同軸上定位。職級越高區間越寬 —— 每個區間的上限假設具備 8 年以上直接相關經驗。 长条长度代表该职位的薪酬跨度，并非在共同轴上定位。职级越高区间越宽 —— 每个区间的上限假设具备 8 年以上直接相关经验。
ML Engineer vs Data Scientist 機器學習工程師 與 數據科學家 机器学习工程师 与 数据科学家
These two titles get mixed up more than any others, and hiring the wrong one costs you a quarter. The distinction is what the person is accountable for at the end of the month. 這兩個名銜最常被混淆，請錯人就是以季度計的代價。分別在於：這個人月底要為哪一件事負責。 这两个头衔最常被混淆，请错人就是以季度计的代价。分别在于：这个人月底要为哪一件事负责。
Machine Learning Engineer
- Accountable for the model running in production為模型在生產環境穩定運行負責为模型在生产环境稳定运行负责
- Writes software: services, tests, deployment寫軟件：服務、測試、部署写软件：服务、测试、部署
- Cares about latency, cost per inference, uptime關注延遲、每次推論成本、可用率关注延迟、每次推理成本、可用率
- Hire when the model already works and must ship當模型已可行，需要上線時請他当模型已可行，需要上线时请他
Data Scientist
- Accountable for whether the answer is correct為「答案是否正確」負責为「答案是否正确」负责
- Writes analysis: statistics, experiments, models寫分析：統計、實驗、模型写分析：统计、实验、模型
- Cares about confidence, causality, business impact關注信賴度、因果關係、業務影響关注置信度、因果关系、业务影响
- Hire when you don't yet know what to build當你還不知道該做什麼時請他当你还不知道该做什么时请他
Not sure which role you need? 不確定你需要哪一個職位？ 不确定你需要哪一个职位？
Describe the problem you are trying to solve — not the job title. We will tell you which role solves it, and which one would be a waste of money. 描述你想解決的問題，而不是職位名稱。我們會告訴你哪個職位能解決，哪個職位只會浪費錢。 描述你想解决的问题，而不是职位名称。我们会告诉你哪个职位能解决，哪个职位只会浪费钱。