But the percentages deserve scrutiny. The underlying studies cover different job boards, collection periods, and definitions. They offer useful signals—not one unified census of what every senior candidate must know.
Start with the denominator, not the headline
Axial Search’s primary research covers 12,148 public U.S. data science postings from January through September 2026. Its methodology combines programmatic rules and AI analysis to extract requirements. Crucially, Axial describes the results as mention rates: an omitted skill is not necessarily an unwanted skill.
That distinction matters when reading KDnuggets’ synthesis. A percentage across all data science vacancies does not establish the requirement rate for senior vacancies. Nor does the proportion of AI-related jobs assigned to senior candidates tell us what proportion of senior jobs requires AI.
A job description is evidence of what an employer advertises—not a complete account of its interview rubric or eventual hiring decision.
Python and SQL remain central, but rankings vary
The two analyses agree on the importance of programming and querying, while differing considerably on frequency:
| Skill or platform | 500-posting analysis reported by KDnuggets | Axial Search’s 12,148-posting analysis |
|---|---|---|
| Python | 74% | 82.8% |
| SQL | 64% | 56.3% |
| Cloud platforms | 62% | 35.0% |
| Microsoft Azure | 28.5% | 18.7% |
| AWS | 19.7% | 26.9% |
These are separate samples, not directly interchangeable measurements. Even the ordering of Azure and AWS reverses. Neither establishes a universal cloud winner for senior data science hiring.
For Microsoft-focused candidates, the narrower conclusion is useful enough: Azure appears materially in both datasets. KDnuggets also reports Power BI in 24.7% of its smaller sample. That supports considering Microsoft’s data ecosystem when evaluating relevant vacancies—not treating it as mandatory everywhere.
Production capability is the stronger common thread
Axial’s leadership framework places hands-on execution and AI literacy above strategic influence. It also reports median requested experience of five years for senior roles, seven for principal roles, and ten for director-level positions and above. Those are advertised experience thresholds, not automatic promotion milestones.
Independent corroboration comes from Andres Vourakis’ FutureProof Data Science research. His October 21 analysis retained 7,756 unique U.S. postings after removing duplicates and reposts from 11,953 collected through Google Jobs between November 2025 and October 2026. Among postings naming an AI skill, 62% also requested production capabilities such as deployment, serving, or monitoring, compared with 31% of those without an AI-skill mention.
The practical interpretation: a polished notebook can demonstrate experimentation. A stronger senior-level portfolio should also explain how the work becomes dependable software.
For an Azure-oriented example, Microsoft Learn’s deployment training explicitly covers MLflow model versioning, GitHub environment approvals, managed online endpoints, automated testing and deployment through GitHub Actions, monitoring, and rollback. That is a concrete illustration of production work extending beyond model training.
AI matters beyond AI-labelled jobs
FutureProof found that roughly half of its postings expected AI skills or AI-project work, although only 13% included AI in the title. LLMs appeared in approximately one-quarter of postings; retrieval-augmented generation and AI agents each appeared in roughly one in ten.
Its seniority breakdown adds an important qualification: specific AI skills appeared in 42% of junior, 42% of mid-level, and 43% of senior postings, increasing to 51% for Staff+ and 57% for technical leads. AI was therefore not exclusively a senior requirement in that dataset.
These observations support learning to build and evaluate AI-enabled systems where target roles demand it. They do not support declaring every traditional statistical or analytical role obsolete.
Certifications need a correction—and a reality check
KDnuggets reports cloud certifications in approximately 20% of its smaller sample. Axial, however, puts its eight most-mentioned certifications together below 2%, with AWS Certified Machine Learning–Specialty at 0.2%. The available evidence does not reconcile that gap. Candidates should not interpret either number as a universal purchasing recommendation.
There is also a concrete Microsoft correction: Azure Data Scientist Associate, associated with DP-100, retired on June 1, 2026. Microsoft identifies Machine Learning Operations Engineer Associate, exam AI-300, as its replacement. A historical job-posting reference to DP-100 is not a current exam plan.
Degrees remain a separate consideration. Axial reports degree requirements in approximately 93% of its postings, reinforcing that evidence of practical delivery and formal eligibility are not necessarily substitutes.
Turn the findings into an application plan
Rather than collecting frameworks like souvenir mugs, use a small, targeted exercise:
- Select ten relevant vacancies. Keep geography, seniority, and industry reasonably consistent.
- Separate required from preferred qualifications. Record degree and experience constraints independently.
- Group requirements by capability. Distinguish data preparation, statistics, deployment, AI implementation, and communication.
- Attach evidence to each recurring requirement. Use a project, technical decision, deployed system, or documented outcome.
- Prepare one end-to-end explanation. Describe the problem, baseline, evaluation, deployment choices, operational limitations, and business result.
This is an application strategy, not a measured guarantee of hiring success.
The central lesson is straightforward: use posting research to identify patterns, then validate those patterns against the jobs you actually want. Demonstrating ownership of a working system is a more coherent story than presenting an alphabetized inventory of tools.