I am a new graduate working as a data scientist at a mid-sized company. I'm wondering about the importance of LeetCode for data science interviews. I'm comfortable with SQL and can handle hard-level SQL problems, but I struggle with medium and hard-level coding questions (non-sql/algorithm style) on LeetCode. On the flip side, I am very comfortable with ML and DL algorithms. Should I focus more on building ML/DL projects or on mastering medium and hard-level LeetCode problems?
LeetCode/DSA-style problems are terrible enough already for software engineers - I imagine it's even worse for Data Scientists. I don't think you should spend significant time studying LeetCode (especially the hard-label questions) unless you are 90%+ sure an upcoming interview will have that kind of question.
I've talked with data scientists a little about their interviews and I've talked with data engineers even more: The consensus seems to be that LeetCode isn't nearly as common for them as it is for SWE while SQL very much is (which makes sense as it's what data folks actually do).
This economy is also one of the absolute worst to blindly grind LeetCode in as 80% of Big Tech companies are doing hiring freeze.
Check out the resources here as well: [Taro Top 10] Effective Interview Prep
In my experience, Leetcode style questions are not common for Data Science interviews. However, for MLE/Applied Scientist/ML Scientist they can occur sometimes.
For me personally, I am working more on the Engineering side, so I am trying to master a baseline level Leetcode competence (not trying to grind leetcode but atleast familiar enough so that if I have an interview with LC, I can spend 1-2 weeks brushing up what I need) by spending 3-5 hours a week on LeetCode
Regarding data science interviews, I recommend this book: Ace the Data Science Interview: 201 Real Interview Questions Asked By FAANG, Tech Startups, & Wall Street