Machine Learning Theory Interview (By Hiring Manager)
1. Interviewer started asking about my thesis and asked questions about IMU sensor, Odometry
2. Classification metrics (Precision, Recall, curves - what is measured along X, Y and how is area calculated, how you conclude about best threshold)
3. Decision Trees and Random forest regularization techniques
4. Advantages of LLM over LSTM
5. Explain LLM architecture and every layer functionalities
6. Scenario-based question: How would you ensure a particular class is distinguished correctly during training and achieves zero errors? (By assigning a higher weight to that class during training.)
7. Explain about Adaboost and how you can modify according to your dataset
8. Scenario based: If the Product Manager provides a new requirement, how would you approach and execute it?
9. Scenario based: Explain a scenario from your prev experience where you had to work with multiple stakeholders
10. Scenario based: You are given a requirement to develop a solution that relies on external AI APIs. However, the Product Manager does not want any sensitive data to be sent outside the system. How would you handle this situation?
I would initially use the external APIs but ensure that all sensitive user data is masked or anonymized before sending any requests. This allows rapid prototyping while keeping data privacy intact. In parallel, I would begin working on developing an in-house model capable of handling the required tasks, so that we can eventually migrate away from external APIs entirely.
(This was the same approach eventually adopted by the team.)
