Ho presentato la mia candidatura online. Ho sostenuto un colloquio presso Infrrd
Colloquio
It was overall pleasant and quite challenging. You must be confident about your machine learning basics. Especially the math behind. Questions about machine learning are quite exhaustive and comprehensive. Try not to be stuck during the interview. If you prepare enough you should be able to pass.
Following a technical assignment was a virtual interview conducted over Zoom, in which I was asked questions about my career goals, Python skills, and domain knowledge for the role I was applying for. I was also briefed about the company and the nature of work along with the responsibilities of the role I was interested in.
Domande di colloquio [1]
Domanda 1
Python function to parse a string and count the occurrences of every word.
Ho presentato la mia candidatura tramite un'altra fonte. Ho sostenuto un colloquio presso Infrrd nel mese di ago 2020
Colloquio
I had no idea I was by default applying for ML intern when I emailed my resume. And more than 2 weeks has passed since the interview and I think it's safe to assume I didn't get the position.
1. HR emailed some prescreening questions like programming languages and expected compensation.
2. An entity recognition challenge to be completed in three days. I was able to solve it and give my analysis (No feedback about this)
3. ML-based technical interview without instructions about contents. I got unprepared bc I was mostly preparing for resume projects and general DS interview questions like what is F1 score, etc.
Domande di colloquio [1]
Domanda 1
We spent some short time in discussing resume projects. But then we spent the rest of time discussing a single ML model in depth. From data formulation, processing to how the model works to regularization methods. So I say you have to at least review the models to a complete graphic level and shallow-ish math equation level in order to answer the questions. (BTW the interviewer will expect a "correct" answer)
How Linear regression works? How are the weights updated? How by adding a regularization term can reduce overfitting? (Think mathematically)
And questions like how to deal with imbalanced labels? (I provided 2 solutions but the interviewer was expecting something else, and honestly we don't learn that in graduate school...)