算法部署与数值精度专家（存算与AI） - 芯片研发
Core
Expert in algorithm deployment and numerical precision for in-memory computing (IMC) AI chips, bridging algorithm requirements with hardware architecture to ensure high-accuracy, low-power inference for large AI models.
Role type
Senior IC algorithm deployment and numerical precision engineer (IMC/AI chip)
Builds
High-efficiency AI chips based on in-memory computing architecture
Domain
Semiconductor hardware design + AI/Deep Learning
Deliverable
production ML models
Required skills
Model quantization (PTQ/QAT), numerical precision analysis, hardware-software co-design, operator optimization, mixed-precision strategies, error localization, in-memory computing architecture knowledge
Preferred skills
Python, C/C++, deep learning operators (Conv, MatMul, Attention), low-precision data formats (INT8/4/2, FP8/BF16), calibration strategies, heterogeneous deployment (GPU/NPU)
Responsibilities
Analyze model structure and precision requirements to decompose into system architecture and toolchain needs; Design data format and precision schemes considering quantization, rounding, and overflow; Implement precision compensation schemes (bias correction, noise injection); Execute model quantization and calibration; Build precision comparison frameworks between software models and hardware results; Design deployment strategies across IMC chips and heterogeneous platforms.
Seniority
Senior, hands-on IC
