Python

How is client data (metadata) enriched?

Approximately how much does it cost to run and maintain a system?

Why did you leave the Pinkman studio?

Do you think a end user (not an agent) can use the assistant directly?

How do you update your knowledge when tariffs and conditions change (in the context of frequent changes in the insurance market)?

What is the architecture of the RAG-assistant system for an insurance broker?

How can you hedge against these risks?

How to simplify working with multiple insurance companies whose data is updated at different intervals?

Why were you laid off from your previous job?

What do you do when the assistant glitches, hallucinates, or responds incorrectly?

How is request routing (classification by question type) performed?

What does the first working version ('not ashamed to show') look like?

How do you prevent losing the context/structure of a document during chunking?

Which audience and scenario will you prioritize first? Why?

Where will you start your 30-60 day plan?

What company are you looking for?

How did [organization] comply with the requirements of 152-FZ when using LLM?

Which LLM should be used for answer generation (GigaChat, YandexGPT, Chinese models)?

What risks do you see (technical, product, domain)?

How do you collect and structure a knowledge base from documents, regulations, rules? How do you solve complex insurance rules with regional exceptions?