Bloomberg AI scientist 纽约挂经



Q1 -
Q2 - (It was a much simplified version of this. Instead of 3, given two strings check if target can be generated using them, not necessarily interleave)

s1 = ‘bloom’, s2 =‘berg’, target=‘bloomberg’
ans = True

s1 = ‘bloom’, s2 =‘berg’, target=‘bloomberg1’
ans = False

s1 = ‘blom’, s2 =‘oberg’, target=‘bloomberg’
ans = True


What is L1 and L2 norm?
What is the difference between SVM and Logistic regression, why SVM?
What is overfitting and underfitting a model? How to identify and correct it?
After these warm ups, he asked ML system design question:

Design a spam email detection system. What if instead you were to identify spam tweets?
The second interviewer wasn’t convinced with my ML system design at all. We both hated each other by the time it ended. 我问他 how will generally Bloomberg evaluate me AI scientist role since it is not just coding. 他说 ‘Based on the questions we ask you in the interview’