The 3rd round is composed of 2 case questions and some general questions about work style, which are quite authentic.
Staff Data Scientist Interview Questions
118 staff data scientist interview questions shared by candidates
What is your motivation for changing from your current job to Trustly?
basics of ML, stats, information theory. bin search, heaps, heapsort
What was the most personal challenges with fellow employees
Asked about my background and past experience
I was given a take home and I respect that Coursea actually gave me something challenging in terms of cosine similarity matrices. I have applied to several of coursera's competitors and theirs were too easy for them to figure out who is better than whom besides comparing resumes for the tiebreaker. So I was able to get past the take home screener and get to the crux of why they were hiring a data scientist in terms of how would you solve the searchability indexing unsupervised? But in terms of practicality, I was providing my answer for how searchability indexing could be hybrid supervised and unsupervised by speech to text merging of the data...again, they were in their "academic world stance" that this problem should be solved unsupervised only and I was super confused because the data scientist who was interviewing me had an academic perspective of problem solving that was out of touch with real life solutions. Do you want to smoke your pipe and come up with some 'fantasy problem' with a 'fantasy solution' or do you want a problem solver who is able to think outside the box and actually come up with a meaningful solution? again, they were hiring someone who could conform to their culture of thinking, which is super restraint instead of someone who could bring a solution to them much better based on the premise of what the problem is. I am not sure if I would be a good fit culturally because I would be bringing them solutions left and right and someone would still complain that one of the restraints were broken instead of realizing my contributions to that company
System design and ML question.
All things related to your past experience and problem solving along those lines.
If training two boosted trees models for new and old users will give different results then training one model on all users and which is better approach.
Why AUPRC had been not improved but AUCROC had been improved when you made the change?
Viewing 41 - 50 interview questions