Data Analyst - Assignment

Thank you for showing interest in Talentwunder and for wanting to be part of Talentwunder's Data Team.

This assignment is a chance for you to showcase your technical skills, your creativity as well as your business sense. We hope it will form a good base for further discussions in the interview process.

Rules

How much time do I have?

We value your time and so should you. Please post your solutions in 7 days from receiving the task.

How much time should I spent on the tasks?

The assignment should not take longer then 6 hours total. However, Time is somtimes hard to find. If you had less time at hand, please tell us. We try to account for it when assessing your solutions.

How should I post my solutions?

The easiest would be to fork this repository and commit your solutions to your fork. Please detail instructions on how to execute your solutions. If you use jupyter notebooks or similar, document your thought process in respective markdown fields. If you develop your own reusable python code in python scripts, we consider it a plus.

Need any more clarification?

If you still have some questions regarding the assignment, please reply to the email we send the assignment in, or contact us directly: career@talentwunder.com

---

1. Programming Language Statistics

The marketing team of Talentwunder wants to write a blog post on one of our supported networks 'GitHub'. They want to educate our users on the types of candidates present at GitHub. To this end, they would like to show statistics on the programming languages that are used by the GitHub users.

1.1 Data

For the purpose of this coding challenge, we have sampled a small portion of our GitHub profiles, which are stored in a JSONL. To obtain this file, please extract data/github_profiles_samples.zip. The resulting github_profiles_samples.jsonl contains one JSON object per line, which corresponds to one GitHub profile each. The profiels have the following structure:

_id: unique identify used by the database
repos: [
    # A list of objects, with structure:
    name: name of the github repo
    stars: number of stars given to this repository
    forks: number of repositories forking from this repository
    commitCount: number of commits pushed for this repository
    contributors: Number of programmers contributing to the code in addition to the creator/author of the repository
    isFork: indicates if this repo is a fork of another
    languages: [
        # A list of objects, with structure:
        name: name of the programming language
        percentage: GitHubs estimate of how much code was written in the corresponding language
    ]
    lastUpdated: Object holding the date when the repository was last updated
]
activity: 
    visitFreqLast90Days: Statistic reported by GitHub how often the user has logged in in the last 90 days.
    numContributionsLast365Days: Statistic reported by GitHub how much commits have been pushed in the last year.

1.2 Tasks

1. Parse the programming languages per profile from the provided file github_profiles_samples.jsonl
2. Compute and visualize statistics on the programming languages, that you would deem useful for our marketing team. E.g.:
   • most frequent programming languages
   • mean number of profiles/candidates per language
   • most frequent combination of programming languages
   • programming language pair correlations
   • ... make your own decision! You do not need chose any statistic suggested here.
3. Bonus: Can you think of a method to group GitHub profiles by their programming languages? For example to distinguish frontend/web developers from backend developers, data scientist, and so on. No implementation required!

---

2. Proficiency Measure

Talentwunder is gathering the data of millions of potential candidates from all over the web. At the same time, we try to make it easy for our users to interpret the massive amount of data that we end up with. For example, we would like to help our customers in assessing the proficiency level of a candidate at a programming language based on his GitHub profile.

2.1 Data

See data description (1.1) of the first task "Programming Language Statistics"

2.2 Tasks

1. Come up with a measure/statistic that approximates the proficiency of a candidate in a specific programming langauge. Proficiency to our users could mean anything from years using the programming language to simply count how much code has been written. Keep in mind your time constraint and settle for some reasonable approach. Bonus: If you have a more intricate approach, feel free to outline that. (No implementation required)
2. Implement a function that consumes a GitHub profile and returns your proficiency measure per language
3. Bonus: Wrap the function in a python script that can be called from command line.

---

3. GitHub Brainstorming

Based on your own knowledge and what you have done so far: Can you think of other ways you can use the provided data? Pick one idea that you have, preferably from the following domains:

• Data visualizations and story telling for marketing
• Profile enrichment, helping users understand and assess the candidate behind the GitHub profile
• Insights about our data for the purpose of business intelligence or job-market insights