分析架构师岗位职责

深度学习性能分析架构师NVIDIA英伟达半导体科技(上海)有限公司,英伟达DeepLearningPerformanceArchitect(C++/C)-Shanghai,Hsinchu

工作职责:

*针对架构和算法特征开发汇编级或者CUDA代码

*优化cuBlas、cuDNN、TensorRT的核心kernel;

*针对未来GPU架构开发原型代码,推进下一代架构的设计和优化;

基本要求:

*严谨的逻辑思维和分析能力

*较强编程能力(C/C++)、算法分析和实现

*熟悉计算机体系结构优先

*有CUDA代码调优经验(或者SIMD等架构的调优经验)优先

*熟悉矩阵计算的优化和加速优先

详细JD如下

DeepLearningPerformanceArchitect-Shanghai,Hsinchu

Areyouobsessedwithperformance?Doyouliketoworkattheintersectionofhardwareandsoftware?Doyouliveandbreathedeeplearning?NVIDIAisseekingworldclassprogrammersandperformancearchitectswholovetosqueezeouteverycycleofperformancefromdeeplearningcodes.Inthisrole,youwillwritecodethatshipsinourdeeplearninglibraries,aswellasguidethedirectionofourfutureGPUarchitectures.Thispositionofferstheopportunitytohaverealimpactinafast-moving,technology-focusedcompany.

Whatyou'llbedoing:

-Developstateoftheart,performancecriticalcodetoacceleratedeeplearningonNVIDIA'splatforms.

-DevelopinnovativeHW,DSP,GPUandsystemarchitecturestoextendthestateoftheartindeeplearningperformanceandefficiency

-Analyzeandprototypekeydeeplearninganddataanalyticsalgorithmsandapplications

-Understandandanalyzetheinterplayofhardwareandsoftwarearchitecturesonfuturealgorithmsandapplications

-Collaborateacrossthecompanytoguidethedirectionofmachinelearning,workingwithsoftware,researchandproductteams

Whatweneedtosee:

-MSorPhDinrelevantdiscipline(CS,EE,Math)

-TrackrecordofoptimizingcodeforperformanceonCPUsorGPUs,includingassemblyorSIMDprogramming

-Strongmathematicalfoundationinmachinelearninganddeeplearning

-ExperienceworkingwithdeeplearningframeworkslikeCaffe,TensorFlowandTorch

-StrongprogrammingskillsinC,C++,Perl,orPython

-FamiliaritywithGPUcomputing(CUDA,OpenCL,OpenACC)andHPC(MPI,OpenMP)

-Strongbackgroundincomputerarchitecture

-Experiencewithmatrixmultiplyandconvolutionalgorithms