Skip to main content
T4 Automotive and self-driving Pony.ai Internship

Pony.ai AI Infra internship

Pony.ai AI Infra internship interview questions covering operator optimization and high-performance computing

Operator optimization High-performance computing

Fundamentals

  1. What is a CUDA Graph? Why does using CUDA Graph use more GPU memory? Which stage of inference is a better fit for CUDA Graph?
  2. Explain the core ideas of the NVIDIA GPU PTX machine model.
  3. Describe the full CUDA compilation pipeline.
  4. What are MLIR’s design goals, and why is MLIR needed?

High-performance computing

  1. Explain cross-Block communication and Warp-level primitives. List common Warp primitives and what they do.

Coding

  1. Implement merge sort.