AI engineer
An employee asks, "How many leave days do I get?" The system checks their permissions, retrieves the HR policy passage and answers with a citation, or says it does not know. See RAG and capstone 4.
Job titles vary between companies, and many jobs mix both. This page shows the centre of each role so you can choose a roadmap. The guide covers both.
| Aspect | AI engineer | ML engineer |
|---|---|---|
| Main question | How do we build a reliable product on top of existing models? | How do we learn a predictive model from our own data and run it? |
| Typical work | RAG assistants, agents, document processing, AI features inside apps | Churn, fraud, demand, pricing, quality and recommendation models |
| Core skills | LLM APIs, prompting, embeddings, retrieval, evaluation, security, APIs | Data preparation, features, algorithms, validation, MLOps |
| Languages | Python or C#; both are used widely | Mostly Python; C# with ML.NET in .NET companies |
| Maths in daily work | Little: metrics and similarity scores | Some: statistics and metrics; tools do the heavy maths |
| Hardware | A normal laptop; models through an API or run locally | A normal laptop for most tabular work; GPUs for deep learning |
| Parts of this guide | Parts 1, 2, 5, 6 and 7, plus the basics of Part 3 | Parts 1, 2, 3, 4 and 7, plus the basics of Part 6 |
| Common confusion | Calling an API is the start, not the job: evaluation, security and operations are most of the work | ML is much more than LLMs; most business ML is on tables of data |
An employee asks, "How many leave days do I get?" The system checks their permissions, retrieves the HR policy passage and answers with a citation, or says it does not know. See RAG and capstone 4.
A model trained on two years of customer records predicts who is likely to cancel next month, is served as an API and is checked for drift every week. See classification and capstone 1.
| Role | What they do |
|---|---|
| Data scientist | Answers business questions with data: analysis, experiments (A/B tests), and models. Overlaps strongly with the ML engineer; usually less production engineering. |
| MLOps or platform engineer | Builds the pipelines, registries, serving and monitoring that let other teams ship models safely (Part 7). |
| Data engineer | Builds and runs the pipelines that collect, clean and store data (Part 2 at large scale). |
| Research scientist | Invents and tests new model designs and training methods; usually needs advanced maths and a research background. |