Behavioral questions
The questions the interviewer asked were fairly simple, roughly like this:
- Give me a story where you have limited time to finish a project
- Give me a story where you actively learn a new skill (Learn and Be Curious)
- Give me a story (where you showcase "Bias for Action")
- Give me a story (where you showcase "Ownership")
I felt like while interviewing me, the interviewer might have been filling out a form where they need to summarize my stories into bullet points. So thinking about it the other way around, the material you prepare for yourself should have those bullet points too — short, but information-dense and directly answering the question (Amazon's leadership principles). Helping the interviewer fill out their form is basically helping yourself.
ML Basics
The interviewer was very nice, and the questions were pretty basic, but I think it really depends on which team you're interviewing with.
- What are GD and SGD, and what's the difference between them? Why use SGD?
- What is batch size?
- What are supervised learning and unsupervised learning? Give some algorithm examples.
- Explain the principles of linear regression / logistic regression from scratch.
- Explain the loss function for logistic regression.
- Explain basic machine learning concepts: learning rate, regularization.
- How do you deal with overfitting?
- Explain bootstrapping and boosting.
- Explain how boosting algorithms work.
- How does reinforcement learning (RL) relate to supervised and unsupervised learning?
- How are rewards used to update a policy?
- How are neural networks integrated into reinforcement learning?
- How is a Transformer different from an RNN? What bottleneck does it solve?
- What is an embedding?
- During the embedding process, how do you distinguish the meaning of "King" in a chess context versus a kingdom context?
- How do you fine-tune a large language model (LLM) when compute resources are limited?
Discussion
Loading comments…