Assessing the fastest-growing job is tricky. For example, using job-posting data isn’t great because there could be lots of duplicate postings out there or a lot of the postings could be going unfilled. Another big issue is defining exactly what a job is: The exact same responsibilities could be associated with the job title “data scientist”, “data engineer” or “ML engineer”, depending on the particular job titles a particular company decides to go with. So, whoever’s evaluating job growth is going to end up bucketing groups of related jobs and responsibilities into one particular, standardized job-title bucket, probably these days in a largely automated, data-driven way; if you dug into individual examples, I’m sure you’d find lots of job-title standardizations you disagreed with but some kind of standardization approach is essential to ensuring identical roles with slightly different job titles get counted as the same thing.