Python

What could the red fields in the EXPLAIN ANALYZE results indicate?

What is GraphQL and how does it differ from REST?

Did your database serve as a cache or was a separate caching tool used?

Is it possible to implement the same functionality as a context manager but without using the 'with' statement? For example, opening a file, working with it, and ensuring it is closed.

Tell about your experience using AI/agents in work (personal projects)

What other ways can be used to solve the problem of parallel insertion of identical data?

What is RAG?

In a table without a primary key, constraints, unique indexes, or relationships with other tables, and with duplicate rows, how to eliminate duplicates and prevent them from occurring again?

How do large language models (LLMs) work, who invented them, and why does it work?

A string is immutable. But if I have a variable, I assign it a string, and on the next line I reassign a different string — it doesn't crash. How can this be explained?

If we created a new column with the BIGINT type and filled it with the same data as the old INT column, how do we handle the fact that two primary keys cannot exist simultaneously?

Transactions have the concept of ACID. Can you name 1-2 principles and explain what they do?

Is the number of decorators on a function limited in any way?

Tell about your work experience (not related to AI) — your last project.

What happens if the same data is inserted into the database simultaneously within one millisecond?

When can a situation occur where the complexity of a dictionary operation is worse than O(1) due to storage features in memory?

Python has GIL — how then does the backend handle multiple concurrent requests?

How does the search for relevant (similar) vectors occur, and what is used under the hood?

I wrote the same function with calculations in C++ and Python, ran it in multithreading. Fast in C++, slower in Python. Why?

We have a microservice with one endpoint that executes only one SQL query. When running locally, the endpoint responds in 1-2 minutes instead of the expected 50-200 milliseconds. How would you troubleshoot this issue?