Interview Practice Questions

Compare pointwise and pairwise ranking approaches in terms of quality, speed, and metrics.

What is the difference between MAP and NDCG?

Name the main challenges in recommendation systems and ways to address them.

What would you do if an A/B test showed no difference in the primary metric, GMV?

Which ranking metrics do you know?

Which neural network architectures for recommendation systems do you know?

What is the difference between SASRec and BERT4Rec?

Which offline and online metrics would you use for recommendations? How would you run an A/B test?

Which metrics are used to evaluate ranking?

How would you collect target labels for a search ranking task?

Do you have experience with transformer models for ranking?

How would you incorporate user session context into ranking, for example when recommending accessories for a product the user has already selected?

How would you optimize ranking for margin or revenue rather than simply the number of sales?

How would you use multimodal product features, such as text and images, to improve ranking?

How does implicit ALS (alternating least squares) work with implicit feedback?

How is NDCG calculated?

Describe your most recent recommendation systems experience: what did you do, how did you do it, and what were the results?

Describe your implementation of a transformer-based recommender, SASRec: data, training, and results.

What is the difference between NDCG and MAP in ranking?

What types of loss functions are used in ranking tasks?