We are seeking a skilled AI Compiler Optimization Engineer to optimize AI model inference performance through advanced compiler technologies. You will focus on performance tuning for CPU or hybrid CPU/XPU heterogeneous architectures, profiling AI frameworks to discover new optimization opportunities, and delivering cutting-edge insights from industry research.
Key Responsibilities:
Compiler-Based Performance Optimization:
- Implement compiler techniques (e.g., MLIR level optimizations, LLVM backend optimizations) to enhance inference performance on CPU and CPU/XPU hybrid systems
- Optimize JIT level compute graphs with operator fusion, memory allocation and etc. for latency/throughput improvements
- Preferred: Experience with LLVM/MLIR development
AI Model Profiling & Framework Optimization:
- Profile end-to-end inference workflows on frameworks like TensorFlow, PyTorch, ONNX, and llama.cpp to identify hotspots and bottlenecks
- Propose and implement optimization strategies (e.g., kernel tuning, graph-level optimizations)
- Preferred: Experience optimizing models on multiple AI frameworks
Research & Insight Development:
- Track and analyze the latest advancements in AI & compiler research (academic papers, open-source projects)
- Produce actionable insight reports summarizing trends, benchmarks, and potential optimizations
- Preferred: Strong technical writing skills with prior publications or reports