Data science challenges with Filtered — How we do it
Filtered Enterprise accounts now give you the power to test and automatically score candidates on their model-training skills in Jupyter Notebook.
This is the first platform of its kind to offer a tool that:
- Runs the entire Jupyter testing experience in the browser,
- Lets candidates work with millions of rows of data,
- Handle scoring objectively and automatically with a rubric system, and
- Adds video questions and explanation sections to add a human touch.
Let’s walk through how simple it is to get this type of test running for your own data science candidates:
Before you interview
Although Filtered has a library of ready-to-go data science challenges, let’s pretend that you want to build a brand new challenge from scratch.
You or the Filtered team would first prepare a data file that the candidate has to work with. Next would come the question prompt, and finally an answer key.
The Filtered team would connect the answer key to the rubric system which would then be able to read solutions as they are submitted and automatically score them.
Last up, the interview itself would need to be created. It’s best to add 2 or 3 video questions to help determine communication skill level and culture fit. You can set a time limit on video answers, and also a time limit on the data challenge itself.
Once the new data science challenge question is attached to the interview, you’re good to go.
The interview workflow
Start by sending candidates an invitation to take your new Filtered data science interview.
Candidates will be given the opportunity to answer the video questions and then will be taken into our proprietary version of JupyterHub to work on the challenge. At this time, candidates can choose to complete the challenge in either Python or R.
Once submitted, you will be notified about a new interview to review. The challenge will be scored by identifying the MAE (Mean Absolute Error) and RMSE (Root Mean Squared Error), the lower the better.
Like all Filtered interviews, you will have the opportunity to review the candidate’s behavior while they were working, the candidate’s social media, the candidate’s video answers, resume, and solution explanation.
As candidates finish their interviews, they will begin to appear in a leaderboard so you can talk to the best candidates first.
Along the way, you may choose to shortlist specific candidates and eventually schedule some for a final-round interview. Filtered has the tools that you need to get the job done.
Chat with us to test your own candidates
Filtered data science interviews are a part of a Filtered Enterprise account. Introduce yourself and we’ll get you set up fast.
Interested in more data science reading? Check out this article from Filtered CEO and co-founder Paul Bilodeau.
