How does sparse retrieval (BM25) differ from dense?
Machine Learning / AI
What is OCR-free document understanding (Donut, Pix2Struct)?
Design a system for reusing embeddings across different teams within a company.
What are the features of product ranking in e-commerce (CTR, conversion, revenue)?
How do you assess your level of working with SQL?
What are the features of working with PII when preparing datasets?
What tasks do GNNs solve in molecular design (AlphaFold relatives)?
What are class activation maps (CAM) and Grad-CAM?
How will you search for a table with data about a client, the company's director, and his field if you don't know where this data is stored and there's no one to ask?
What is Great Expectations and what are its key concepts?
What main features did you use for spam classification in leads?
What is pre-LN vs post-LN warmup and convergence issues at the start of training?
Why was Recall maximized if you say that you did not try to hit all?
What are typical bottlenecks in LLM serving?
What are the methods of automatic broken pipeline detection?
What is prefix caching in inference?
Tell about the project for evaluating dispatcher performance.
What data quality tools (Great Expectations, Soda) are you familiar with?
What is Linformer and why does it achieve linear complexity?
Briefly tell us what you did with models. What is your experience in ML projects?