What AI engineers do, and what employers ask for
LinkedIn's Jobs on the Rise 2025 report ranked AI engineer as the fastest-growing job in the US. The title covers several jobs, so look at what the postings require.
What 895 job posts asked for
Alexey Grigorev's AI Engineering Field Guide analysed 895 "AI Engineer" postings from 590 companies (builtin.com; Los Angeles, New York, London, Amsterdam and Berlin; collected in early February 2026). 69% were AI-first roles (building AI products), 29% AI-support roles.
| Skill | Share or count of postings |
|---|---|
| Python | 82.5% |
| RAG | 35.9% (the most common GenAI pattern) |
| Prompt engineering | 29.1% |
| LLMs (named explicitly) | 25.4% |
| TypeScript | 23.4% |
| LangChain | 18.8% |
| Agents | 14.4% |
| Cloud: AWS / Azure / GCP | 359 / 214 / 205 postings |
| Docker / CI/CD / Kubernetes | 277 / 262 / 260 postings |
| PyTorch | 197 postings |
| Vector databases | 97 postings |
| FastAPI | 96 postings |
What that means for you
- It's an engineering job first. 93.1% of AI-first roles require skills beyond GenAI: Python, APIs, cloud, containers, CI/CD.
- Evaluation is a differentiator. 39.6% of AI-first roles explicitly ask for evaluation skills, and the guide calls evaluation the single biggest differentiator in take-home assignments.
- Some ML is expected. 64.3% of AI-first roles ask for some ML knowledge; fine-tuning appears in 30.8% but is a primary responsibility in only 4%.
- Frameworks come and go. Learn the concepts (retrieval, tools, evals), then any framework is a week's work.
These posts are from the US and Europe. Your local market may differ, which is why this module's task is to collect your own data.
Three kinds of AI companies
Janvi Kalra's advice (quoted in the guide): sort the market into product companies (they use models to build products), infrastructure companies (tools for AI developers) and model companies (they train models), then target the one that excites you. Most AI engineer jobs are in the first group.