Loft Data Scientist Interview Questions
& Process
Real candidates share what happened, how many rounds they had,
and how the experience turned out.
Based on 9 interview experiences · FREE TO READ
Candidate interview experiences
First-hand accounts from people who interviewed at Loft.
Loft
Data Scientist
Mid Average Positive experience No offer 4 rounds
Interview process
Take home Technical screen Presentation Background check Offer
Interview formats
Technical Presentation Case
The interview process has 4 stages: online test, interview with data scientists, a case study, and an interview with HR and leadership. I participated in the first three, but didn't pass the case study. For the case study, you get the problem description and training/test data, and have one week to deliver. After that, there's a 1-hour call to present your results to 3 data scientists. A positive aspect is that they provide feedback on your code and study suggestions. The only negative is the process duration, around 2 months total.
Confirmed questions1 question
- Why did you choose this particular model/metric?
Loft
Data Scientist
Entry Average Positive experience No offer 4 rounds
Interview process
Take home Technical screen Presentation Background check Offer
Interview formats
Technical Case Presentation Behavioral
I really liked Loft's selection process, I found it very transparent. The process is divided into 4 stages: - Online technical test (30 True/False questions about the area) - Profile interview (around 30 min, simple questions about your experience, why you applied etc) - Case presentation (they give you a week to do it, in the pdf they detail what they want and what should be presented, it was very challenging and at the same time fun to do the case, in the presentation they question you why you did it this way, why you used this algorithm etc) - Final interview, I didn't pass the case presentation, but this interview would be just to talk to HR, they would give me a proposal etc As I said, I didn't pass the selection process, but on the same day as the case presentation, they already sent me an email saying I didn't pass and even gave me great feedback on my case, positive points and areas for improvement. I thought that was very cool of Loft, many times companies don't even respond to the candidate or even though they already have the final answer, they prefer to wait a few days to respond. I decided to keep studying and try again in a few months, the company seems great!
Confirmed questions1 question
- Did you analyze what you were removing before removing outliers?
Loft
Data Science Manager
Manager Difficult Neutral experience No offer 3 rounds
Interview process
Recruiter call Technical screen Take home
Interview formats
Behavioral Technical Coding
The first stage was a 'ask me anything' style interview in Brazilian Portuguese with a data science manager. I made it to the second stage, scoring between 60-70%. The second stage consisted of a multiple-choice test on machine learning. The third stage was an open coding challenge to develop a machine learning model. The first interview was good for understanding Loft's data science structure. However, the second stage isn't inclusive and doesn't support diversity as Loft claims. It has 40 questions on Codility to be answered in 30 minutes, less than a minute per question. Even if you answer all correctly, the max score is 75%. The questions are more academic and rarely used in industry, potentially tricky if you're experienced but not recently out of a Master's or PhD. Although Loft is Brazilian and operates in São Paulo and Rio, the Codility questions were in English. If you know Brazilian academic ML but not English ML concepts, you'll be disadvantaged. To pass, strong English and recent academic ML background from abroad is recommended. Alternatively, team up with friends who are in ML Master's/PhD but lack industry experience.
Confirmed questions8 questions
- Regarding OLS, what are the criteria for it to be a Blue Linear Unbiased Estimator?
- For Regularized Linear Regression, if a practitioner wants a sparse parameter vector, should L2 regularization be preferred over L1?
- Is the Mean Absolute Error (MEA) less sensitive to outliers compared to the Mean Squared Error (MSE)?
Loft Data Scientist Interview Questions
Quoted word for word from Loft interview reports.
“Is KL divergence a symmetric measure of dissimilarity between probability distributions?”
Read reports →“Is an autoencoder with a single hidden layer and linear activation functions equivalent to PCA?”
Read reports →“Does DBSCAN operate under the assumption that densely packed samples belong to the same cluster?”
Read reports →“Is the VC-dimension of a k-Nearest Neighbors model with k=1 greater than that of any linear regression model?”
Read reports →“Do convolutional layers and max-pooling layers contribute to making neural networks invariant to translation?”
Read report →“Can Bayesian linear regression, when using informative priors on the weights, be viewed as a form of regularization?”
Read report →“Is K-means a specific instance of a Gaussian Mixture Model where the covariance matrix is diagonal and constant?”
Read report →“In the context of reinforcement learning, will an epsilon-greedy policy with a constant epsilon exhibit linear regret?”
Read report →“For a convex loss function, is stochastic gradient descent likely to require fewer iterations for convergence than standard gradient descent?”
Read report →Formats, difficulty and experience
Across all 9 Loft interview reports.