Show HN: I collected 500 python & SQL questions from data science interviews
platform.stratascratch.comVery cool. I like the SQL questions.
I have a question I'm going to throw out there for HN. How many of you bounce back and forth between SQL and Python, even though you could stick within SQL alone? For example, one question (Airbnb) was about year-over-year growth by looking at the count of hosts registered for each year. I might be inclined to go as far as getting the count through SQL, but then (maybe in a notebook) just write the rest in python.
I'm not asking how many people use SQL and Python together (my guess is almost everyone does), I mean how often do you use SQL just for the select, join, filtering and aggregation, but then go to python even though you know there's a way to do this without leaving SQL?
In practice, rarely ever. I do my data pulls and some joining and manipulation in SQL. But anything more complicated gets done in python.
I enjoy the challenges, however, I find it really frustrating when I lose my work every-time I accidentally navigate back to the previous page. This happens easily with a Macbook Pro's trackpad. This heavily detracts from usability and prevents me from 'wanting to complete' the rest of the challenges.
I recommend two modifications:
1) Disable Navigating Forwards/Backwards in the browser, like Jupyter Notebook
2) Make use of HTML5's localstorage / HTML Web Storage for each challenge.
Hi, thanks letting me know. I'll make sure to add that feature in ASAP. Appreciate it.
One AirBNB SQL question prompt is:
"Find the search details made by people who searched for apartments designed for a single-person stay."[0] This could be interpreted to mean some refined version of the following:
select * from airbnb_search_details WHERE accommodates = 1 AND property_type LIKE 'Apartment';
However, the expected output includes non-apartments, for example Houses. The expected output matches the following query instead:
SELECT * FROM airbnb_search_details WHERE accommodates = 1 AND beds = 1;
What do you think users should take away from the experience of completing this question?
0: https://platform.stratascratch.com/edu-content-editor?id=961...
Yea, I completely understand what you're talking about. There's many "solutions" because sometimes there's no real solution. In practice and in an interview, often times it's about a dialogue and aligning with an approach and assumptions.
My implementation of having 1 solution isn't good enough. I plan on adding a discussion forum to discuss these exact topics.
Hey guys, just wanted to share a resource that I created a while back but recently redesigned and relaunched. I collected 500+ data science interview questions for SQL and python coding questions. I kept many of the questions as is and re-wrote others while still testing the same concepts they would test for in interviews. The platform has a fully executable IDE with datasets. Many of the questions are free. I'd love it if you guys could share your thoughts on it! Does this hit the mark as an educational resource to prep for interviews? What can be improved?
Really nice, I have a csv of something similar for Java oriented questions but I really enjoy the work you've done here as I'm increasingly asked questions about everything since going full stack
Thanks so much. The data science interviews are a bit of the kitchen sink as well. I am also planning to release a bunch of technical non-coding interview questions that cover statistics, modeling, machine learning, product sense, business cases, etc. The biggest problem I'm facing is that it's taking me forever to write up solutions for these types of questions. I'm thinking about making it a discussion forum so others can help contribute.
But thank you for the kind words. I appreciate you taking the time to look at it.
Anytime, it's a cool resource to see :) Maybe some people out there don't agree with charging for this type of information, but I think the fact that you've also extended the $25 offer to anyone that has an .edu email address is a lot better than those who charge $600 for something pulled off of a Leetcode discussion forum ;)