Screening Round
Questions:
-
ML Integration: Asked to build a model in a Jupyter notebook environment, given a data set.
- Had to build the target variable from the data and drop unrelated columns.
- Built a model that performed better than random using balanced class accuracy due to class label skew.
- Follow-up question on probabilities; explained AUC curve.
- Allowed to check syntax for various tasks, e.g., probabilities from logistic regression classifier.
- Emphasis on building something end-to-end rather than perfect.
-
Coding Round: Multi-part question, Leetcode easy, mostly straightforward. Not in a typical Leetcode style.
Candidate's Approach
For the ML integration round, I focused on understanding the data and identifying the target variable. I ensured to drop any irrelevant columns to improve model performance. I utilized balanced class accuracy to handle the skew in class labels and explained the AUC curve during the follow-up. In the coding round, I approached the multi-part question methodically, ensuring I covered all parts and emphasized writing comprehensive test cases.
Interviewer's Feedback
No feedback provided.
Onsite Round
Questions:
-
Coding Round: Multi-part Leetcode easy question (successfully solved 2 parts) with an emphasis on test cases.
-
Bug Squash: Given a Hackerrank link with a trained model and a code package that loaded the model and ran tests. Identified and fixed 1 of 2 bugs, stumbled at the last step of fixing the second bug. Allowed to reference API docs from the web.
-
ML Design Round: Non-standard design question specific to Stripe. The interviewer frequently interrupted for clarifications, preventing completion of the design. Had to provide a brief overview at the end of the hour.
-
Manager Chat: Standard Sr-level questions on ambiguity handling, mentoring junior folks, and continuous learning. Focused on quantifying impact, but lacked concrete revenue increase numbers for my project.
Candidate's Approach
In the coding round, I ensured to address each part of the question thoroughly and focused on writing robust test cases. During the bug squash round, I quickly identified both bugs, fixed one, but faced challenges with the second. I utilized available resources like API documentation to assist in my debugging process. For the ML design round, I aimed to outline my thought process but struggled due to interruptions. In the manager chat, I discussed my experiences but noted the need for more concrete metrics regarding project impacts.
Interviewer's Feedback
No feedback provided.