So I had my 1st round final interviews for c3.ai today! It was an OK experience.
The interview consisted of 3 portions, and I thought it was a pretty good interview process!
First interview: Let’s say you are working for Expedia. How would you build a model to predict which trips are being taken for business and which are being taken for leisure?
Second interview: Questions on domain knowledge of ML. Questions involve: 1. AUC/ROC. If we have AUC of 0.1? If we have AUC of 0.5? If we have AUC of 0.9? 2. Vanishing Gradient: how to address? 3. Different metrics for evaluation 4. Addressing Dataset Imbalance. 5. How exactly is L1/L2 regularization performed in XGBoost? 1. We can add this as a function on our objective function: we minimize the loss, but also some regularization term too! 6. How would we deal with missing data, when it is a categorical variable! (I actually mostly nailed this question!)
Third interview: Code up minesweeper. Straight up.
I only got through the initialization part of this question (i.e. generating the board, and placing the mines). But I found that despite this, I had the chance to clearly communicate my understanding of CS to the interviewer. The code I wrote for the portion I completed ran perfectly, and the interviewer told me that no one ever finishes the entire minesweeper in the time allocated.
Some reflections and areas where I can improve my data science skills/process:
One area that I am proud of: I have enough domain knowledge, and expertise, to talk “around” questions. Even if I don’t know the direct answer to a question, I can easily answer in the “general” area of a question, which is often sufficient for the interviewer.
One thing I want to note: I am feeling extremely dismotivated, and just feel like relaxing all the time, after getting my Microsoft and Lyft offers. I have been eating a lot, and just wasting time for wasting time’s sake these days. In short, I am ascended. However, this is a bad thing in reality! Writing is a great outlet to get the creative juices flowing, and get me to do things! In the future, I want to reduce the turnaround; the loss on these days will be computed as quite large! (For instance, we can compute the loss in the following manner: 1. 2 hours (of COH); this was not really needed. Also, we could have done something more productive with this time, like looking up something (playing around with RL, etc.) 2. 3-6 hours yesterday, once we got the Lyft and MS offers. )