Airbnb Data Science Challenge Fail
I applied to a data science position at Airbnb last month, and failed at the data science challenge project, although I thought I did really well and invested a lot of time. Can someone at Airbnb tell me what you guys typically look at in a submission? Do you compare the f-1 score against those of other candidates and choose the best like in Kaggle competitions? was I supposed to use complex ensemble modeling techniques? thanks!
Not at Airbnb, but I'm at a FAANG and have passed the Airbnb take-home before. I feel like with these, the last thing they would do is compare f-1 scores. As long as you demonstrate that you're able to do a decent job with the problem and show good practices, comparing metrics is really useless. Some of the other things that are important include: - clean code - good code commenting - ability to explain clearly and concisely what you did and why you chose the things you did - ability to talk about some of the tradeoffs between your choices and other things you could have done - good ability to tie the algorithm/data used to a real business use case and verbalize how you improved the business
I see, thanks! Did you pass your take home when you interviewed there? And so basically you’re saying that it’s ok to limit myself to standard models (boosted trees, random forest, LR, SVM, etc.)?
I did pass the take home, but I did a different track than you (I'm guessing you were on the Algorithms track). But I definitely didn't do anything super fancy and I'm fairly sure it's not a requirement to use anything beyond standard models.
It’s not only about story telling, but also about finding obvious pitfalls in the input data, making reasonable assumptions and explain them well, and creating useful latent features to feed into your model.
I see. But I had never done this before so I’m not sure what kind of pitfalls I should have been looking for
Like imbalanced missing data, extreme outliers, unreasonable records—e.g. end_date > start_date, etc.
These challenges always frustrate the living hell out of me. Such a commitment when you have a full time job.
I don’t mind, because theoretically I can demonstrate my talents in it more than in a phone interview. I just want to learn for next time I apply to the same company or somewhere else with data challenges.