Interview Practice Questions

What methods of communication between microservices do you know?

What are the advantages and disadvantages of a microservice architecture?

How do you deploy a model to production? Is there a separate ML environment?

How are models deployed to production?

Describe your system design experience, specifically with neural network architectures. Have you designed a reasonably large architecture from scratch, or have you mainly adapted existing architectures?

Here are the high-level requirements for the system we would like you to design. The domain should be familiar to you: used-car sales, one of the company's core products. Design, and preferably draw, a recommendation system architecture that meets these requirements.

How would you generate realistic synthetic packet timestamps while preserving monotonicity?

How to assess ML task complexity during project planning?

Will pre-computed embeddings be stored in cache?

Strategy shows good results on small capital. What will happen when scaling to large capital?

Come up with a simple feature-indicator for mean reversion strategy and for momentum

What happens when customer stops topping up balance in fintech products?

How will all of this work in production?

Given a 2D grid (image), a starting point (sr, sc), and a new color, implement flood fill: recolor the starting pixel and all connected pixels of the same color (4-directional).

Write a class for weighted random sampling. The class accepts an array of (item, probability) pairs, and its sample() method returns an item according to that distribution.

Implement weighted random sampling with O(1) query time using the Alias Method.

Implement sampling from a discrete distribution specified by a histogram that may contain floating-point weights. Optimize it to O(log n) per query.

For each number in array B, find the number in array A that maximizes their XOR. Implement a naive solution, then optimize it.

Write a function that, given the mean of five numbers and one of those numbers, returns the mean of the remaining four.

Implement a Bloom filter. Explain how it works and write the code.