Who are we?
We are the System Architecture Group at StarFive in Singapore. We focus on micro-architectural design, modeling, exploration and benchmarking. We help leadership and different stakeholders to make a data driven design decisions.
Job scope:
- Understand the mainstream GPGPU architecture and their design choice
- Using existing ML or HPC background to propose the possible architecture improvements and exploration possibilities
- Can deeply understand the programming model of each architecture and its impacts and influence on eco-system
- Maintain and build the confidence model for both verifying with software tools and profiling the performance so at to understand different architecture design choices
What are we looking for?
- BS/MS/PhD with relevant experience in Domain Specific Accelerators/Machine Learning
- Knowledge of different machine learning algorithms (eg: CNN, LSTM, DNN, GNN etc) and their execution model in GPU/CPU; understanding of parallel execution models like SIMD, SIMT etc
- Familiarity with GPGPU (eg: CUDA/OpenCL) programming models is preferred but not mandatory
- Any prior experience in developing simulation-based performance models for domain specific accelerators
- Proficient in C/C++ and scripting languages (Perl/Python)
- Knowledge in one or more AI programming frameworks – Tensorflow, Pytorch etc
- Knowledge of RISC-V ISA is valuable but not mandatory
- Ability to deliver high quality analysis/results and independently drive modeling tasks is preferred
- Curiosity and Enthusiasm to explore advance state of the art technologies with calculated risks
- Strong interpersonal skills, written and oral; good team player
