Skip to contentSkip to footer
  • Community
  • Jobs
  • Companies
  • Salaries
  • For employers
      Notifications

      Loading...

      Elevate your career

      Discover your earning potential, land dream jobs, and share work-life insights anonymously.

      employer cover photo
      employer logo
      employer logo

      Loft

      Engaged employer

      About
      Reviews
      Pay and benefits
      Jobs
      Interviews
      Interviews
      Related searches: Loft reviews | Loft jobs | Loft salaries | Loft benefits
      Loft interviewsLoft Data Science Manager interviewsLoft interview


      Glassdoor

      • About / Press
      • Awards
      • Blog
      • Research
      • Contact Us
      • Guides

      Employers

      • Free Employer Account
      • Employer Centre
      • Employers Blog

      Information

      • Help
      • Guidelines
      • Terms of Use
      • Privacy and Ad Choices
      • Do Not Sell Or Share My Information
      • Cookie Consent Tool
      • Security

      Work With Us

      • Advertisers
      • Careers
      Download the App

      • Browse by:
      • Companies
      • Jobs
      • Locations
      • Communities
      • Recent posts

      Copyright © 2008-2026. Glassdoor LLC. "Glassdoor," "Worklife Pro," "Bowls" and logo are proprietary trademarks of Glassdoor LLC.

      Company Bowl sample

      Want the inside scoop on your own company?

      Check out your Company Bowl for anonymous work chats.

      Bowls

      Get actionable career advice tailored to you by joining more bowls.

      Followed companies

      Stay ahead in opportunities and insider tips by following your dream companies.

      Job searches

      Get personalised job recommendations and updates by starting your searches.

      Top companies for "Compensation and Benefits" near you

      avatar
      CBRE
      3.5★Compensation and benefits
      avatar
      Jones Lang LaSalle
      3.5★Compensation and benefits
      avatar
      RE/MAX
      3.5★Compensation and benefits
      avatar
      Greystar
      3.8★Compensation and benefits

      Data Science Manager Interview

      28 Mar 2021
      Anonymous interview candidate
      No offer
      Neutral experience
      Difficult interview

      Application

      I applied online. The process took 2 weeks. I interviewed at Loft in Mar 2021

      Interview

      Primeira etapa é uma entrevista pt-br no estilo ask me anything com um gestor de data science Segunda etapa é uma série de perguntas de múltipla escolha sobre machine learning A terceira etapa seria um desafio aberto de código para desenvolver um modelo de machine learning A primeira entrevista é bem legal e dá pra conhecer melhor a estrutura de data science da Loft. Eu cheguei até a segunda etapa, com uma nota entre 60%-70%. A segunda etapa não é inclusiva e realmente não visa a diversidade, da maneira como a Loft se propõe. A segunda etapa consiste de 40 perguntas para responder em 30 minutos na plataforma Codility. Ou seja, pra cada questão você tem menos de um minuto pra responder. Se você dedicar 1 minuto para responder cada pergunta e acertar todas que você respondeu, sua nota máxima será de 75%. As perguntas são mais voltadas para um ambiente acadêmico sobre coisas que raramente você precisará usar na indústria. Portanto, se você terminou o mestrado ou doutorado há um tempo e tem mais ênfase e experiência na indústria, as perguntas podem ser tornar capciosas. A Loft é uma empresa brasileira, e hoje, só atua no mercado de São Paulo e Rio de Janeiro. No entanto, as perguntas do Codility foram todas em inglês. Se você domina machine learning num contexto acadêmico brasileiro, mas não tem domínio dos conceitos também em inglês, então você será passado para trás. Se você quiser muito passar da segunda etapa, sugiro que você tenha um bom domínio do inglês e tenha saído recentemente de um mestrado ou doutorado no exterior com um ótimo conhecimento acadêmico. Outra alternativa é fazer o questionário em dupla ou trio com alguns amigos nerds sem experiência na indústria que ainda estejam num mestrado ou doutorado relacionado machine learning.

      Interview questions [19]

      Question 1

      What are the criteria for an OLS to be Blue Linear Unbiased Estimator?
      Answer question

      Question 2

      In Regularized Linear Regression, if the practioner desires a sparse parameter vector, is L2 regularization preferable over L1 regularization?
      Answer question

      Question 3

      MEA is less sensitive to outliers than MSE?
      Answer question

      Question 4

      If we multiply all probas by 2, the area under ROC curve will be divided by 2?
      Answer question

      Question 5

      Computation of the silhouette coefficient scales well, since its time complexity is O(n)?
      Answer question

      Question 6

      DBSCAN assumes densely packed samples balong to the same cluster?
      Answer question

      Question 7

      Bayesian linear regression, putting informative priors on the weights can be see as a form of regularisation?
      Answer question

      Question 8

      KL divergence is a symmetric dissimilarity function between distributions?
      Answer question

      Question 9

      The entropy of a Normal(0,1) distribution is larger than the entropy of a normal(0, 10) distribution?
      Answer question

      Question 10

      For an SVM with polynomial kernel, reducing the degree of polynomial can prevent underfitting?
      Answer question

      Question 11

      K-means is a special case of Gaussian Mixture Model with a diagonal and constant covariance matrix?
      Answer question

      Question 12

      An autoencoder with one hidden layer and linear activations is equivalent to PCA?
      Answer question

      Question 13

      Comparated to OLS, GLM can assume different erro distributions, such as Poisson?
      Answer question

      Question 14

      t-SNE is a dimensionality reduction algorithm based on matrix factorization?
      Answer question

      Question 15

      Convolutional and Max-pooling layers help NN be translation invariant?
      Answer question

      Question 16

      The VC-dimension of a KNN k=1 is larger than that of any linear regression?
      Answer question

      Question 17

      In reinforcement learning, an epsion-greedy policy with constant epsilon will have linear regret?
      Answer question

      Question 18

      For a convex loss, stochastic gradient descent is likely to take less iterations to converge than gradient descent?
      Answer question

      Question 19

      The predictions of a Decision Tree fitted using the L1 loss are taken by the mean of its leaves?
      Answer question
      1