Skip to main content
T5 Others Transsion New grad Round 1

Transsion AI Infra campus recruiting first round

Real Transsion AI Infra campus recruiting first-round interview questions covering inference optimization and operator optimization

Inference optimization Operator optimization

C++ basics

  1. Difference between pointers and references
  2. Differences between heap and stack memory
  3. Causes of wild pointers and how to use smart pointers

Model deployment and inference

  1. Basic principles of object detection algorithms
  2. Impact of class count and dataset size in object detection tasks
  3. What are YOLOv5’s main improvements?
  4. How to choose a quantization calibration dataset
  5. Basic principles and implementation of quantization
  6. How do you evaluate accuracy after quantization?
  7. Common methods to compute the quantization scale
  8. How to compute inference throughput
  9. Comparison and selection among common inference frameworks

Operators and compute graphs

  1. How compute graphs are built
  2. Low-level implementation of convolution operators
  3. Blocking optimization strategies for matrix multiplication
  4. Basic concepts and uses of the ARM NEON instruction set

Large models

  1. How a large-model tokenizer works
  2. Role and implementation of KV Cache