Банк вопросов из реальных собеседований

A query like "book" returns a million relevant products. How not to waste the whole million with a heavy model?

What model files, metrics, and signals are stored after launch to replicate a decision and notice a change in the format of a particular bank?

You need high-quality sentence embeddings for semantic retrieval. Which data, positive and negative pairs, and loss functions would you use?

If the new carousel was not yet in the product, how to assess the relevance of its recommendations before the experiment and not take the absence of the old display as a negative label?

How do you determine the dataset string, time partition, and metrics of the price model or recommendations so that the offline valuation reflects the actual application?

How to arrange a platform where the main search for candidates is considered a batch, and the operational layer changes the order according to the fresh actions of the user?

How to get training pairs of queries and relevant road segments with a limited manual markup budget?

How to improve the order of the found road segments with a more accurate model and notice the degradation of the search in time?

The bank receives PDF statements from other banks. How to turn documents into verifiable compliance signals?

Transactions and the state of the exchange glass are given. How do you define a target, build signs, do a temporary check, and choose the first basic solution?

In a large archive of trips, you need to find rare traffic situations by text request. How to determine the user, search unit, metrics and the first version of the system?

Which user, product, cart, and interaction features would you feed into a boosted-tree or MLP ranker for cart recommendations?

LLM extracts features from a PDF report: for example, a future production plan. How can you check that a feature is based on a document and not on external knowledge or guesswork?

How should the production system process finished calls, write bookings and avoid operational issues such as double-booking a slot?

After the baseline and ranker, how would you design production serving: candidate retrieval, feature storage, and updates when the cart changes?

The report said: production will grow by 20% for the year, growth will begin in the second half of the year. The model needs a quarterly forecast. What should the LLM feature return?

How would you choose the A/B assignment unit, estimate the MDE, and validate the experiment before exposing users to dynamic delivery pricing?

In the call center, you need to select a security question to verify the client: fairly secure, but not too complicated. How to build a ML question ranking system?

Большая temporal model хорошо ловит события, но слишком дорогая по latency и compute. Как сжать ее для production?

The ISP wants to warn users about phishing pages. How to design a ML detection and warning system?

The agent confidently reports incorrect information. What engineering measures reduce the risk of fictional facts and dangerous activities?

Tell us about the modern LLM architecture and learning process: what are the main stages, data, objective and loss used?

How does speculative generation work and under what conditions does a large number of proposed tokens actually accelerate the underlying model?

The e-commerce application launches a TikTok-like video feed on the main page. There are products attached to each video, there are about 1500 videos and they live for 1-2 months. There is no history on the new surface. How to design a recommendation system?