AI Compiler Intern, CUDA Tile IR - Summer 2027
Core
Intern role focused on improving the performance and expressivity of CUDA Tile IR, a new way to program NVIDIA GPUs for deep learning workloads.
Role type
Deep learning compiler engineering intern
Builds
CUDA Tile IR language constructs and compiler optimization algorithms
Domain
Deep learning, GPU architecture, compiler engineering
Deliverable
production ML models
Required skills
C/C++ programming, software design, debugging, performance analysis, test design
Preferred skills
MLIR, LLVM, XLA, deep learning models and algorithms, deep learning framework design, open-source contributions, CPU/GPU architecture knowledge, CUDA/OpenCL programming
Technologies
CUDA Tile IR, MLIR, LLVM, XLA, CUDA, OpenCL
Responsibilities
Analyze deep learning networks, develop compiler optimization algorithms, add new constructs to the language, expand support to new hardware generations, collaborate with deep learning software framework and hardware architecture teams
Seniority
Intern
