AI Compiler Optimization Engineer

Role Overview

We are looking for an experienced AI Compiler Optimization Engineer to enhance the performance of AI model inference by leveraging advanced compiler technologies. This role primarily focuses on optimizing performance for CPU and hybrid CPU/XPU heterogeneous computing environments. The engineer will also be responsible for profiling AI frameworks to uncover new optimization potentials and translating the latest industry research into actionable improvements.

Key Responsibilities

• Apply compiler optimization techniques, such as MLIR level improvements and LLVM backend enhancements, to boost inference efficiency on CPU and CPU/XPU hybrid platforms.
• Enhance JIT compute graph execution through operator fusion, improved memory allocation, and other strategies aimed at reducing latency and increasing throughput.
• Conduct profiling of complete inference workflows within popular AI frameworks including TensorFlow, PyTorch, ONNX, and llama.cpp to detect performance bottlenecks and hotspots.
• Design and execute optimization measures, including kernel-level tuning and graph-level improvements across multiple AI frameworks.
• Monitor the latest research in AI and compiler technology through academic literature and open-source development to inform optimization approaches.
• Create detailed reports offering insights into emerging trends, benchmarking data, and viable optimization strategies.

Preferred Experience

• Proficiency in LLVM and MLIR development environments.
• Extensive experience optimizing AI models across multiple frameworks.
• Strong technical writing abilities demonstrated through prior publications or technical reporting.