Master's or Doctoral degree in Computer Science or Electronic Engineering
5+ years of experience in low-level computing algorithm development
Experience with AI accelerators or large scale parallel computing systems
Deep understanding of workload characteristics for large language and multimodal models
Familiarity with AI software stacks and algorithms.
Responsibilities
Build an accurate AI performance model for theoretical analysis
Conduct surveys on emerging hardware designs and analyze cutting-edge technologies
Identify performance bottlenecks in AI workloads with the research team
Define algorithmo-hardware co-design features for next-gen chips
Model performance for AI workloads with various training and inference algorithms
Lead team in developing acceleration algorithms for efficiency trade-offs
Track and analyze trends in algorithm-hardware co-design technologies
Benefits
Permanent position with immediate start
Opportunity to work with cutting-edge AI technologies
Collaborative work with an experienced AI research team
Engagement with industry-leading hardware designs
Potential for significant contributions to the future of AI chip design
Full Job Description
Huawei Canada has an immediate permanent opening for a Principal Scientist.
About the job:
Build an accurate and universal AI performance model based on mainstream AI acceleration technologies to support theoretical analysis.
Track the emerging hardware designs in the industry, conduct in-depth insight and survey analysis, and identify the direction of key cutting-edge technologies.
Cooperate with our AI research team to identify key performance bottlenecks in future AI workloads, and define key algo-hw codesign features of our next-generation chips, for the objectives of low cost, high throughput, great scalability, and stability.
Performance modelling of representative AI workloads with state of the art training & inference algorithms on different hardware specs for quantitative analysis of compute, memory, IO and interconnect.
Lead our team for acceleration algorithm breakthrough in best tradeoff between model quality and compute efficiency.
Track the emerging algorithm-hardware codesign technologies in the industry, conduct in-depth insight and survey analysis, and deeply understand main directions and trends of cutting-edge algorithm-hardware codesign technologies.
About the ideal candidate:
Master's or Doctoral degree in Computer Science or Electronic Engineering.
At least 5+ years of experience in low-level computing algorithm development, AI accelerator/ large scale parallel computing / high performance computing system design is an asset.
Deep understanding of the basic principles and workload characteristics of large language models / multimodal models, the popular AI software stack (operators, compilers, acceleration libraries, frameworks) and mainstream large model training and inference algorithms, such as hybrid parallelism, low precision data formats, sparsity, P/D splitting, etc.
Familiarity with microarchitecture of AI chips is an asset.
