About the position
The position involves designing software systems, including layout and flow charts, tailored for common neural network frameworks aimed at BST AI SOCs (computer chips). The role requires defining scalable neural network development tools and applications to facilitate effective mapping and optimization of network models. Additionally, the candidate will be responsible for developing automated model conversion flow features that are user-friendly and generic, suitable for most common models that end users might create. The position also includes developing tools and applications to retarget various hardware architectures supported by BST AI SOCs, including CONV, GEMM, EDP, and DSP. Furthermore, the candidate will develop tool and application flows to support post-training quantization and quantization with or without retraining, as well as automate the optimization search process. The role requires executing moderately complex software test plans, which include software functionality, integrity, and security, based on industry tools and practices, departmental guidance, and policies. The candidate will also perform coding using C++ or other programming languages.
Responsibilities
- Design software systems including layout and flow charts for common neural network frameworks aimed at BST AI SOCs.
- Define scalable neural network development tools and applications for effective mapping and optimization of network models.
- Develop automated model conversion flow features that are user-friendly and generic for common models.
- Create tools and applications to retarget various hardware architectures supported by BST AI SOCs, including CONV, GEMM, EDP, and DSP.
- Develop tool and application flows to support post-training quantization and quantization with or without retraining.
- Automate the optimization search process through developed tools and applications.
- Execute moderately complex software test plans, including functionality, integrity, and security testing based on industry standards.
- Perform coding using C++ or other programming languages.
Requirements
- Proficiency in C++ or other programming languages.
- Experience in designing software systems and flow charts for neural networks.
- Knowledge of neural network frameworks and their optimization.
- Familiarity with hardware architectures such as CONV, GEMM, EDP, and DSP.
- Experience in developing automated model conversion features.
- Understanding of post-training quantization techniques.
Nice-to-haves
- Experience with AI SOCs and their development tools.
- Familiarity with software testing methodologies and tools.
- Knowledge of optimization algorithms for neural networks.
