I applied online. The process took 2 months. I interviewed at Chronograph in Jul 2023
Interview
There are 4 rounds with 5 segments
1. HR phone screening: general phone screening questions plus three technical questions (HR just read it aloud)
2. 30-min meeting with the hiring manager: half a deep dive into the resume; half questions related to machine learning practice and problem solving
The next two segments are scheduled together
3.1 30-min coding assessment: in Python; hand-craft function to calculate tf-idf
3.2 30-min Paper presentation: read one of the paper they give (from arXiv) and present its content as well as your comments
4. 90 Virtual on site: Mostly behavioral and cultural fit questions; first half chatting with team lead in neighbouring teams (Software, Data engineer); second half with CTO
You have a dataset of 1000 features. You build a baseline model with all features but you are not happy with the performance. You think you need to reduce the features to improve performance, so what methods you can use (to do feature reduction or selection)?
You have developed a model with appropriate training, validating, and testing. You are about to deploy the model but you still worry that your model overfits. What would you do to check for overfitting?
Clarification: by this stage, you still have data not used in the process above
What is the purpose or advantage of doing A/B testing before deploying the models?
Clarification: by A/B test it means the control variable is: with the model or without the model
Preliminary phone interview with a Recruiter, followed by a Zoom Call with the hiring manager. Very pleasant and conversationally-focused interview process. While I did not receive an offer, the recruiter was able to provide some post-interview feedback, which I appreciated.
Interview questions [1]
Question 1
What experience do you have with fine-tuning various models?
