Some Machine Learning Interview Prep

After extensively preparing for machine learning engineer interviews, I wanted to share my experience and pass on potentially valuable knowledge for others also pursuing a career in ML.

High level taxonomy for “Machine Learning”

First off, to roughly gauge how well you know the terminology, I would recommend taking this test. Leave a comment on what you got! I scored a 23/40. My memory is a bit fuzzy on some terms but according to their website, I beat the mean! Anyway, this website has a lot of good content for refreshing yourself on all things ML.

Some general takeaways:

• a categorical feature can be nominal or ordinal — a nominal feature does not have order (rom-com vs documentary) whereas an ordinal feature has an implicit order (4 star rating vs 5 star rating)
• overfitting occurs typically when the model grows too complex — eg. a tree model has too many levels, the coefficients of a linear model get too high, etc.
• overfitting causes the model to not generalize well and could cause the model predictions to have high variability
• on the flip side, underfitting causes the model to not learn the full picture (they are biased)(see variance bias tradeoff)
• for text processing (also known as NLP), it’s important to stem words (eg. change “playing” to “play”), standardize words (eg. change acronyms to the actual word), and get rid of stop words (eg. remove “what” from all text)
• for classification, you can use metrics such as precision, recall, log-loss, and accuracy as evaluation metrics — take the time to understand each

Second, I wanted to compile some common questions that I encountered regarding various ML related situations. My goal is to provide real examples of where ML can be put to good use. Feel free to grab a pencil and paper and write down what you’d say and the solutions you’d propose.

1. How do you test whether a new credit risk scoring model works? What data would you look at?

2. Automate some portion of the customer support experience.

3. Given some order fulfillment centers, how would you go about improving the efficiency of the fulfillment centers?

When you’ve finished, feel free to take a look at some of the things I decided to touch upon at the bottom of this notebook.

Hope you learned something! Good luck~