J'ai postulé via un recruteur. Le processus a pris 2 semaines. J'ai passé un entretien chez Point72 (New York, NY)
Entretien
Two 30 minute phone screens, and then a take home project to create a predictive model of 311 calls based on weather. I was told that people typically spend 4 to 6 hours on the project and that most applicants are brought in for an interview.
The dataset was several GB, rather messy so it was clear that the 4-6 hours to download, clean, understand the data and create a predictive model was going to lead to something rather rudimentary.
I ended up spending about double that amount of time on the model, my solution wasn't perfect, but it had the basics - data cleaning, basic data exploration, test/train split, out of sample validation. It certainly seemed reasonable given the time parameters that were suggested.
Nevertheless, I did not get invited for an onsite. After I spoke with my recruiter about the process, I found found out that the Point72 recruiter was quite new to the company, and the expectations were much higher than was initially conveyed to me.
Large take home projects are very unfair to candidates since they require a substantial time investment on the part of the candidate, and very little on the part of the firm. If I had known the time commitment expected, I would have declined the take home project.
Questions d'entretien [1]
Question 1
How would you create a predictive model of y given x, where x is between 0 and 10, is continuous and linear on the integers.
Followup: If you were to model this is 10 separate regressions, what would be the effect? If you average the regressions at the integers would these be biased estimates?
Hackerrank consisting of easy-medium SQL and coding leet code type questions followed by "behavioral interview" that is technical, followed by technical statistics round. Interviewer dove deep into my projects. Statistics round was straightforward (bayes rule etc.)
J'ai passé un entretien chez Point72 (New York, NY)
Entretien
One of those pymetrics tests + wunderlic. Online Assesment, 1 SQL (easy), 2 leetcode mediums. Then virtual onsite with 3 interviewers. 1 Behavioral, 1 data case, 1 machine learning high level concepts.
Questions d'entretien [1]
Question 1
What happens to F-Statistic/T-statistic if you replicate the data?
1) 30 min call with recruiter 2) 90 min OA 3) 2 interviews with Data Scientists (1 behavioral + 1 technical) 4) week long case study + create a presentation 5) (onsite round) I was ghosted after my case study submission
The process is long. If you’re not 100% committed to the org + team, I wouldn't invest the time in the case study + presentation round. That said, the interviews themself were fair if you have a solid background in DS
Questions d'entretien [1]
Question 1
What do you do when you have conflicting priorities?