Nvidia Senior Compiler Architect MLIR - LLVM 

United States, Texas 

889632368

What you will be doing:

• Provide technical leadership to a small team of engineers working on compiler middle-end optimizations.

• Analyze the performance of application code running on NVIDIA GPUs with the aid of profiling tools.

• Work with hardware architects, assembly language teams to help define architecture evolution, informed by application use-cases.

• Identify opportunities for performance improvements in the LLVM based compiler optimizer.

• Interact with Open-source MLIR/LLVM community to ensure tighter integration.

• Work with geographically distributed compiler, architecture and application teams to oversee improvements and problem resolutions.

• Be part of a team that is at the center of deep-learning compiler technology spanning architecture design and support through higher level languages.

What we need to see:

• M.S or Ph.D. in Computer Science, Computer Engineering, or related fields (or equivalent experience)

• 10+ years experience in Compiler Optimizations such as Loop Optimizations, Inter-procedural optimizations and Global optimizations and Program Analysis.

• 3+ years technical leadership experience

• Excellent hands-on C++ programming skills.

• Excellent understand of architecture features used in parallel programming (fence, vote, synchronization directives etc.)

• Experience writing significant analysis or transformation passes in LLVM framework.

• Understanding of Processor ISA (GPU ISA would be a plus).

• Strong background in software engineering principles with a focus on crafting robust and maintainable solutions to challenging problems

• Good communication and documentation skills and self-motivated

Ways for you to stand out from the crowd:

• Experience in developing applications in CUDA or other parallel programming language.

• Deep understanding of parallel programming concepts.

• Prior experience working with architecture teams on parallel processing hardware.

• Experience working on compile-time improvements in JIT compilation contexts.