Hiring challenge for the senior MLE (Machine Learning Engineer) applicants at Radix.
We give you four challenges below from which you:
- Choose one challenge for which you design and implement your solution. You are free to choose the format (repository, Colab notebook, ...), but note that we would like to review your code.
- From the remaining three, choose two challenges for which you describe how you would tackle them. The format in which you describe your solution is up to you (plain text, diagrams, etc.).
Note that, for the implemented solution, we don't expect you to attain the perfect result. What is more important is that you show how you take a structured approach to the chosen problem and that you can implement a working solution.
You will come and present all three (one worked out and two prepared) of your solutions at our office once finished. You will be evaluated on the performance and designs of your solutions, as well as on your presentation skills.
We have generated a household energy consumption dataset. In this dataset you will find energy consumption time series of 1000 households. Each time serie spans one month and the measurements were taken with an interval of one hour. The data was generated by different types of households, both in terms of composition and activity patterns. Can you detect them all?
In this video, segment and classify the Smurfs. We provide you with a dataset of segmented Smurfs to help you with the segmentation. After segmentation, classify the Smurfs in the following categories:
- Papa Smurf
- Smurfette
- Vanity Smurf
- Smurf (i.e. the other, regular Smurfs)
Note that we do not provide you a dataset of labeled Smurfs. You will have to find a way to cope with this.
Here you find an illustration of what we'd like to see. Can you smurf them all?
As machine learning engineers, we occasionally stumble across a exiting paper and we might want to look for other similar papers. Given this corpus of NeurIPS paper abstracts, can you build a system that allows to retrieve similar papers given the abstract of another paper?
Classifying movies based on their synopsis is a classic machine learning problem. Let us extend this problem by considering movie posters as well. Can you use the posters to improve classification results? When you choose to implement this solution you can do the evaluation with the Mean Average Precision at K of the top 5 predicted genres.
You can find the data here:
Send your code and solutioning to your contact person at Radix and we'll get back to you with our feedback.