Python is on 21,874 candidate profiles on MyInternships.in. NumPy is on 8,496 and Pandas on 8,056. Put differently, the count for each data library is well under four-tenths of the count for the language it runs on.
That gap is the most useful thing we can tell an employer hiring for data work. "Python" on a profile says very little about what the person does with it. A library name says rather more.
How the library counts compare with the language
Our base is the 55,557 candidate profiles that listed skills on 1 Oct 2026. NumPy sits on 15.3% of them and Pandas on 14.5%. Relative to the Python count, NumPy comes to 38.8% and Pandas to 36.8%.
The same profile is tallied separately for every skill on it. We have not measured how many of the NumPy profiles also list Pandas, though the closeness of the two counts in every city suggests the labels travel together.
| Where | Python | NumPy | Pandas | NumPy as share of Python |
|---|---|---|---|---|
| All India | 21,874 | 8,496 | 8,056 | 38.8% |
| Bengaluru | 2,458 | 914 | 861 | 37.2% |
| Hyderabad | 1,968 | 767 | 751 | 39.0% |
| Pune | 1,778 | 782 | 736 | 44.0% |
| Chennai | 1,550 | 565 | 548 | 36.5% |
| Delhi | 1,162 | 496 | 485 | 42.7% |
| Mumbai | 899 | 362 | 359 | 40.3% |
Pune and Delhi have the densest data-library signal
Bengaluru has the most NumPy profiles in absolute terms. Relative to its Python count, though, it is slightly below the national figure. Pune is the stand-out: its NumPy count is 44.0% of its Python count, the highest of the large cities. Delhi follows at 42.7%.
Chennai is at the other end, at 36.5%. That does not mean Chennai lacks data candidates. It has 565 NumPy profiles, more than Delhi. It means a larger part of Chennai's Python-listing pool describes itself without the data libraries.
If you are comparing where to open a data internship, the ratio tells you how much sifting a "Python" search will need in each city, and the absolute count tells you how many people are there at the end of it.
Why a library name is a stronger filter
A language can be learnt for many purposes: college coursework, competitive programming, scripting, web development or analysis. A library is learnt to do a specific kind of work. Someone who lists Pandas has, at minimum, been pointed at tabular data.
That is why we suggest treating the two library names as a proxy for data-readiness when sorting applications. It is a proxy only. But it is better aligned with an analyst role than the language tag.
It also explains a common frustration. Employers who search for Python and expect analysts often find that many of the results have never loaded a dataset. Searching on the library trims the pool to under half and makes it far more relevant.
There is a caution on the other side. A candidate can list both libraries after one guided course, and the two names tend to be learnt in the same week. The tag narrows the field; it does not rank the people within it.
Where the funnel narrows next
Below the libraries, the counts drop again. Machine Learning is listed by 3,281 candidates and Data Analysis, as a phrase, by 3,390. Both are well under half the Pandas figure.
For most business analytics roles this is reassuring. The work is usually cleaning, joining, summarising and explaining, which is library-level work. You do not need to restrict yourself to the smaller modelling pool unless the job truly involves building models.
Questions that test real data handling
Since the tag is self-declared, follow it with a practical check. These prompts work well in a short screen and need no elaborate setup:
- Given a table with missing values in one column, what are your options and how would you choose?
- How would you combine two tables that share a key but have different numbers of rows?
- Describe a time your summary looked wrong and how you found the cause.
- When would you reach for a spreadsheet instead of code?
The last question matters more than it looks. Excel is listed by 16,153 of our candidates, and a good analyst knows when the simpler tool is the right one.
Where possible, hand over a small, real extract from your own data with anything sensitive removed. Candidates respond better to a concrete business question than to a textbook dataset, and their answers will tell you how they would handle your actual work.
What the figures do not establish
All of this describes labels on MyInternships.in profiles, not assessed skill and not the national workforce. A candidate may use Pandas daily and not think to list it, and another may list it after a single exercise. The comparison is still useful because it shows how differently the three labels behave as search terms.
Hiring data-ready candidates
Put the library names in the JD alongside the language, describe the dataset the person will work with, and ask for one notebook or analysis they can explain. Our employer resources cover JD wording, and recruiter intelligence has more on analyst supply.
Ready to open the role? Hire data interns on MyInternships.in and screen on the work sample, not the keyword. You can see how candidates view such roles on the internships page.


