J'ai postulé via un recruteur. Le processus a pris 6 jours. J'ai passé un entretien chez Floy (München, Bavaria) en déc. 2023
Entretien
After an initial talk with the recruiter, I was invited to a technical, 1-hour interview with the company's head of AI. This interview comprised a discussion about an applicant's old project and then a round of general technical questions. The interview went relatively well, until the interviewer didn't understand a component of my project, assumed he knew better and began trying to explain to me my own work. A second discouraging moment occurred later when he casually mentioned I would get to work *unpaid* overhours on a normal basis, if I would get the job. He stated normal working weeks in Floy have a 60-hour duration and openly asked if I was willing to work at that rate for the same base salary than a 40-hour week (normal for German standards).
After this interview, I didn't receive any other notifications. I find the interview tone and lack of communication extremely impolite. It can only leave you wondering about the work dynamic inside the company.
Questions d'entretien [2]
Question 1
Pitch me one project you have worked with recently.
The worst job interview I've ever had: The first interview was a 30-minute phone conversation. The interviewer came 20 minutes late for the call. The second interview consisted of a coding task to apply machine learning to specific medical data. They asked me to work more than 60 hours per week without paying compensation. There are strict legal limits on working hours in Germany: employees must not work more than 48 hours per week. They want you to break the legal limit.
Questions d'entretien [1]
Question 1
Floy is a medical AI company on the mission to maximize human healthspan. Our first AI product helps radiologists to improve the diagnostic accuracy for lumbar spine examinations. Diagnostic errors are frighteningly common and statistically affect everyone of us in our lifetime. Radiologists have a 26.1% error rate of clinically relevant findings and 38% of these errors can be prevented through the collaboration of radiologists and AI. It is unacceptable that existing technology is not adopted in radiology. It is time to change that!
Our goal is to find out how you approach problems, structure ML projects and obtain tangible results under time constraint. Code quality isless important in this exploratory challenge (production code is a different story). Build a minimal working pipeline with the framework (Tensorflow, PyTorch, etc) of your choice.
1) Structure your ML pipeline. Use different cells with comments to explain you approach.
2) Develop a working pipeline. It does not have to be perfect. Focus on getting things done.
3) Answer three qualitative questions in simple words.
