What is context length and why is it difficult to extend?
Machine Learning / AI
What prompt compression techniques do you know (LLMLingua)?
What is overfitting and how to fight it?
What are the causes of overfitting?
What methods can be used to improve a model?
What is autocorrelation and partial autocorrelation (ACF, PACF)?
How to estimate the median response time of a service with a large volume of logs that do not fit into memory?
How to choose data for median estimation: sequential pass or random subsampling?
What are the types of regularization and how do they work?
What is an experience replay buffer and why is it needed?
Why can model weights become large and why is it necessary to reduce them with regularization?
What is calibration and why is it important for fairness?
What is t-SNE/UMAP visualization of feature space in CV model?
How to organize containerization and deployment of models?
What are Medusa and multi-token prediction heads?
How is the problem of finding a pair of numbers with a given sum solved?
What types of errors occur in medical protocols and how can they be classified?
What is NCCL and what is it responsible for?
Design an OCR + NLP system for processing scanned contracts.
How to test an LLM agent (deterministic tests, replay, mocked tools)?