AI Skills Employers Ask For
Skill frequency across the tracked AI job market: which skills appear most often and in which roles.
Generated Sep 29, 2026 · Dataset window: Dec 5, 2009 – Sep 29, 2026 · Observations: 3,302
Executive summary
Skill demand measured across 3,302 active AI job postings. Skills are extracted from full job-description text, not titles.
Most requested skills
| Skill | Jobs mentioning | Share of all tracked jobs |
|---|---|---|
| Python | 875 | 26.5% |
| Evaluation | 725 | 22.0% |
| LLM | 652 | 19.7% |
| AWS | 583 | 17.7% |
| AI Agents | 560 | 17.0% |
| Google Cloud | 507 | 15.4% |
| Azure | 498 | 15.1% |
| SQL | 482 | 14.6% |
| System Design | 473 | 14.3% |
| Spark | 350 | 10.6% |
| Java | 274 | 8.3% |
| Kubernetes | 260 | 7.9% |
| Observability | 244 | 7.4% |
| TypeScript | 241 | 7.3% |
| Fine-Tuning | 179 | 5.4% |
| C++ | 173 | 5.2% |
| JavaScript | 143 | 4.3% |
| Rust | 128 | 3.9% |
| Guardrails | 124 | 3.8% |
| React | 119 | 3.6% |
| Docker | 106 | 3.2% |
| MCP | 91 | 2.8% |
| PyTorch | 89 | 2.7% |
| RAG | 83 | 2.5% |
| MLflow | 69 | 2.1% |
Methodology note
Each skill is detected by conservative pattern matching against job titles and description text, with one evidence snippet stored per match. Skills below roughly 1% of tracked jobs are omitted from tables to avoid noise.
Methodology & limitations
Dataset: active job postings collected from public employer career pages and ATS APIs (Greenhouse, Lever). Role classification uses exact-title pattern matching with evidence stored per posting; skills are extracted from description text.
Limitations: the dataset covers tracked employers only and is a lower bound on the market. Figures describe the tracked dataset, not the entire AI labor market. Salary figures use advertised ranges only.
Full details: data methodology.