Job description preview
From DeepX’s original posting
Frontier of On-device AI Semiconductors for Everyone, Everywhere Role Overview The DEEPX NPU (DL) Compiler Team develops a deep learning compiler that generates binary code for DEEPX’s proprietary NPU, designed to accelerate inference of deep learning models. This NPU binary code includes crucial optimization features such as computational performance tuning, power efficiency, and resource minimization (e.g., memory optimization). It is one of the core technologies of the DEEPX AI Processor Solution. To achieve this, the Compiler team collaborates closely with the NPU Hardware team to define and implement SW/HW specifications and interfaces. The NPU Compiler development process includes: □ Generating an intermediate graph-level representation from AI software platform code □ Converting DNN models for deployment on DEEPX NPUs □ Applying various optimizations to enhance performance per watt □ Generating NPU binaries considering hardware resource constraints Responsibilities □ Develop deep learning compiler optimized for DEEPX AI products □ Implement various compiler optimizations: tiling, memory optimization, operation fusion, etc. □ Collaborate with the NPU HW team to design and…
Open full posting on DeepX’s site