What are the skills needed for AI engineer?

The skills employers ask for are visible in the postings, not in the course catalogues. Of 26,929 live AI-related postings, 4,357 name at least one specific tool, 70 tools in all. OpenAI leads with 1,062 mentions. The top 20, counted this month, is below, with what each one means for an AI engineer.

The number this article rests on: 4,357 of 26,929 live postings name a specific AI tool, and the top 20 tools account for 7,580 of those mentions. Measured September 2026.

A skills list written by a school tells you what is teachable. A skills list counted from postings tells you what is hired. This page is the second kind. Level reads every live posting for the tools it names, keeps the ones that appear in at least one posting, and ranks them by frequency. The ranking is re-counted monthly, so a tool that fades from postings fades from this page.

What tools do AI engineer jobs ask for?

The 20 most named tools across 4,357 live postings are led by OpenAI (1,062), Claude (989), and Cursor (745). Model providers and coding assistants outrank training frameworks: OpenAI appears 2.1 times as often as PyTorch. The market hires people who build on models more than people who train them. Measured September 2026.

1. OpenAI: 1,062 live postings.
2. Claude: 989 live postings.
3. Cursor: 745 live postings.
4. Clay: 534 live postings.
5. PyTorch: 514 live postings.
6. Databricks: 489 live postings.
7. Claude Code: 418 live postings.
8. Anthropic: 354 live postings.
9. Copilot: 299 live postings.
10. ChatGPT: 297 live postings.
11. Gemini: 270 live postings.
12. Gong: 232 live postings.
13. LangGraph: 225 live postings.
14. LangChain: 189 live postings.
15. Codex: 184 live postings.
16. TensorFlow: 180 live postings.
17. Decagon: 173 live postings.
18. Vertex AI: 157 live postings.
19. MLflow: 138 live postings.
20. Sierra: 131 live postings.

Each link opens the live list of postings naming that tool, so you can check the count against the jobs behind it.

Do AI engineers need to know machine learning frameworks?

Less often than the title suggests. The training-side tools, PyTorch (514), Databricks (489), LangGraph (225), and LangChain (189), are named less often than the model APIs and coding assistants, OpenAI (1,062), Claude (989), Cursor (745), and Claude Code (418). Framework skills still matter for the 206 machine learning engineer postings and the 772 Level 4 roles. Measured September 2026.

The practical split, from the postings:

• Applied AI engineer. Builds products on top of hosted models. Postings name OpenAI, Claude, an orchestration layer such as LangChain or LangGraph, and a coding assistant. 221 postings carry the title AI engineer.
• Machine learning engineer. Trains, evaluates, and serves models. Postings name PyTorch, TensorFlow, Databricks, MLflow, or Vertex AI. 206 postings carry the title.
• LLM engineer. Sits between the two: fine-tuning, evaluation, retrieval. 63 postings carry the title.

Pay is close across the split. Median maximum salary is $274,550 for AI engineer postings (94 with a USD range) and $273,500 for machine learning engineers (74). Pick the side by the work you want, not the pay.

Do I need a degree to be an AI engineer?

Mostly no, by what employers write. Only 34 of 221 live AI engineer postings state a degree requirement, and across the whole catalogue 4,781 postings do (17.8%). What postings ask for instead is named tools and shipped work. Measured September 2026.

That does not make the job easy to enter. Seniority is known on 19,442 postings, and among those 12,056 are senior or lead against 960 junior or intern. The skills list above is the shortest path in: one hosted model API, one orchestration library, one coding assistant, and something you built with them that a hiring manager can open.

Are AI engineers still in demand?

Yes, on the numbers. 221 live postings carry the AI engineer title, 206 machine learning engineer, and 63 LLM engineer. Counting by work rather than title, 772 postings sit at Level 4, where building AI systems is the job, at 369 companies in the catalogue. Measured September 2026.

Demand is concentrated where the models are built and served: Software Engineering (373 Level 4 postings), Data (135 Level 4 postings), and Sales (70 Level 4 postings). The rest of the catalogue hires people who use AI rather than build it, which is a different skills list and a different article: What type of jobs are in AI?.

The jobs right now

Generated from the live catalogue: Level 4 postings, where building AI systems is the job. Newest first, refreshed every six hours.

The full list is on the Level 4 page. For a single tool, open its hub, for example Claude or PyTorch.

Method

• Scope: live postings in the Level catalogue. A tool counts once per posting when its name appears in the text; a posting can name several tools.
• Sample: 26,929 live jobs, of which 11,953 publish a salary and 10,606 publish it in USD.
• Measured: September 2026, from the live catalogue at jobsbylevel.com, refreshed every six hours.
• Currency: USD only, no conversions. Medians are on the posted maximum unless stated.
• Not measured: skills that postings describe without naming a product (Python, statistics) are not counted; seniority is unknown on 7,487 postings.