Applied Scientist Interview Questions

1,159 applied scientist interview questions shared by candidates

Problem: Top K Most Similar Documents You are given: an integer array queryEmb of length D, representing a query embedding a 2D integer array docEmbs of size N x D, representing N document embeddings an integer k All embeddings are already L2-normalized. The cosine similarity between two normalized vectors is equal to their dot product. Return the indices of the k documents with the highest cosine similarity to queryEmb, ordered from most similar to least similar. If k > N, return all document indices sorted by similarity. Function Signature def topKSimilar(queryEmb: np.ndarray, docEmbs: np.ndarray, k: int) -> np.ndarray
avatar

Applied Research Scientist

Interviewed at Microsoft

4
Mar 21, 2026

Problem: Top K Most Similar Documents You are given: an integer array queryEmb of length D, representing a query embedding a 2D integer array docEmbs of size N x D, representing N document embeddings an integer k All embeddings are already L2-normalized. The cosine similarity between two normalized vectors is equal to their dot product. Return the indices of the k documents with the highest cosine similarity to queryEmb, ordered from most similar to least similar. If k > N, return all document indices sorted by similarity. Function Signature def topKSimilar(queryEmb: np.ndarray, docEmbs: np.ndarray, k: int) -> np.ndarray

- Tell me why you're interested in this domain. - Tell me about relevant projects you've done in this domain. - What implications would a multi-modal generative AI tool have (compared to a unimodal one) and how would you go about mitigating its biases. - Case Study: A company wants to use an AI tool to summarize managers' feedback for employees. How would you "systematically" approach identifying the biases of such a tool?
avatar

Senior Applied Scientist

Interviewed at Boston Consulting Group

4.2
Feb 13, 2025

- Tell me why you're interested in this domain. - Tell me about relevant projects you've done in this domain. - What implications would a multi-modal generative AI tool have (compared to a unimodal one) and how would you go about mitigating its biases. - Case Study: A company wants to use an AI tool to summarize managers' feedback for employees. How would you "systematically" approach identifying the biases of such a tool?

Viewing 941 - 950 interview questions

Glassdoor has 1,159 interview questions and reports from Applied scientist interviews. Prepare for your interview. Get hired. Love your job.