Skip to main content
T2 AI unicorns StepFun Internship

StepFun AI Infra internship

Real StepFun AI Infra internship interview questions covering operator optimization, high-performance computing, and related topics

Operator optimization High-performance computing

High-performance computing

  1. Explain DMA versus RDMA: principles and differences.
  2. Why pin memory for cudaMemcpy?
  3. Core difference between processes and threads?

Training optimization

  1. How is sequence parallelism implemented? What was communication overhead in experiments? What further optimizations exist?
  2. Common schemes for training stability and fault tolerance?
  3. Which frontier techniques or research directions in Infra are worth watching?

Coding

  1. Given start/end times of several TV shows, how many shows can you fully watch at most (interval scheduling).
  2. Implement an LRU Cache.