Xiaomi · 北京
端到端算法工程师实习生
Listed on Xiaomi ’s official careers site. Posted April 8, 2026. Check that it is still open and apply there.
Apply on Xiaomi ’s careers site →Job description
- Develop quantization, sparsity, pruning, and distillation techniques to enhance the production-level autonomous driving models.
- Optimize, convert, and deploy autonomous driving models (e.g., ONNX models) that operate efficiently on diverse hardware (GPU, CPU, in-house AI ASIC).
- Design and implement custom kernels using C++ and CUDA to accelerate model operations and pre/post-processing pipelines.
- Perform a systematic benchmarking, scaling, and validation of inference performance across various hardware platforms (GPU, CPU, in-house AI ASIC).
- Collaborate with hardware, compiler, and AI infra engineers to achieve efficient and accurate AI model inference.
Qualifications
- Familiar with techniques like quantization, sparsity, pruning, and distillation for edge and real-time inference.
- Strong proficiency in Python and C++, and deep learning frameworks such as PyTorch.
- Hands-on expertise with CUDA programming, low-level performance profiling, and compiler-level optimization.
- Strong understanding of computer systems and architecture, with experience deploying AI models on GPUs and NPUs.
- Experience with model optimization tools like ModelOpt/ONNXSim/TensorRT, as well as deploying models on edge devices or mobile platforms.
- Strong problem-solving skills with the ability to debug and optimize high-performance inference workloads.
- Experience collaborating with hardware/compiler/AI infra engineers to connect model-level and system-level optimization.