Breadth - generative models, loss functions, tokenizers, LLMs, convolutions, residuals Depth - my research, applications and relation to other fields Math - implement SGD for a problem (derive gradients, update, momentum etc.). Derive closed-form solution to this problem.
Describe some criterion for model selection? Why is dimension reduction important?
You could be asked to describe one of your previous projects, answer behavioral questions, and implement an algorithm from scratch.
Hypothesis Testing on difference of mean explain OLR simple coding question involving calculating some statistics from a csv file
Focus on a relevant project you've worked on and explain it in great detail (specifically spoke about a computer vision task).
"So, what is the main benefit of your [project] method?" "Why might you use ultrashort laser pulses in industry?" "Where might you use lasers in a defence capacity?"
For the technical portion, was asked how to design a dictionary from scratch without using a dictionary. Discussed options such as storing data in a fixed array, or sorting into a list. The best solution uses a hashmap and open addressing to handle possible collisions.
1. Formulas of cost functions
Any difference between transformer architecture and that of LLM
what is the difference between overfitting and underfitting
Viewing 271 - 280 interview questions