Ho presentato la mia candidatura online. Ho sostenuto un colloquio presso Bloomberg (Londra, Inghilterra)
Colloquio
I interviewed for Bloomberg AI, London, UK. Interviewed till the final rounds. The whole process was two rounds of 1-hour video interview (one round per day) + Full day interview with 4 rounds.
It's very gender imbalanced. All interviewers that I met (about 10 people from AI team) were all male.
Domande di colloquio [1]
Domanda 1
Video interviews were LC coding + machine learning questions + previous research/projects technical questions (for example, you used X in a project. explain algorithms of X. what loss function did you use?)
Full day interviews were similar to the video interviews. But one of the rounds was very software engineering focused.
LC questions were medium level, parenthesis, data streaming, k-largest.
Ho sostenuto un colloquio presso Bloomberg (New York, NY)
Colloquio
(1) Recruiter phone screen
(2) Hiring manager 1 hr interview, with another group member shadowing
(3) 1 hr presentation of past work
(4) 4 hr virtual on-site with the team
Domande di colloquio [1]
Domanda 1
Given categorical and numerical dataset, asked about ML system design. Later write code to implement some functions.
Ho presentato la mia candidatura tramite segnalazione di un dipendente. Ho sostenuto un colloquio presso Bloomberg nel mese di gen 2022
Colloquio
There were 3 rounds in total. Two phone screens and then one virtual onsite. The phone screens consisted of conceptual ML theory questions and a Leetcode type question. For onsite, there were 3 technical rounds (one focussing on leetcode + ML theory, one focussing on only leetcode, and one focussing on ML design + leetcode) + HR round + Manager interview.
Domande di colloquio [2]
Domanda 1
ML Questions:
Q. Write an ML algorithm for a step function (explain loss, hyperparameters, etc)
Q. You have a dataset of demographic features and want to do income group prediction, what will be the preprocessing, model, metrics, etc?
Q. How does the Transformer work? How does attention work?
Q. What are vanishing gradients? How do you solve them?
Q. Design an ML system to separate different threads of conversations in a single group of multiple people
Leetcode Questions:
Q. Given an input stream of words, print top k words (with the highest frequency) at any time.
Q. Given an input buffer of processes with different start and end times, print them in the order of their end times.