Skip to main content
T0 Big tech ByteDance New grad

ByteDance AI Infra campus hire (1)

ByteDance AI Infra campus hire interview questions covering inference optimization

Inference optimization

Inference optimization

  1. What common methods accelerate large-model inference?
  2. What model compression and quantization techniques exist? What are the main benefits of quantization?
  3. What gains come from weight quantization, activation quantization, and KV Cache quantization respectively?
  4. How does KV Cache work? Why does PagedAttention help?
  5. How does Prefix Cache work, and which scenarios fit it?
  6. Building an AI Agent app, which core modules do you roughly need?

Project experience

  1. Dig into your project experience.

Coding

  1. A ring has 10 nodes (labeled 0-9). Starting from node 0, each step you move one node clockwise or counterclockwise. How many distinct walks return to node 0 after n steps?