AI Compiler Optimization Engineer

Our client is looking for a talent-driven AI Compiler Optimization Engineer to elevate AI model inference through state-of-the-art compiler techniques. In this role, you will lead performance tuning across CPU and hybrid CPU/XPU heterogeneous systems, profile deep learning frameworks to uncover new acceleration opportunities, and translate emerging academic research into impactful engine enhancements.

Key Responsibilities

Compiler & Code Optimization

• Apply advanced compiler transformations (such as MLIR optimizations and custom LLVM backends) to boost execution speeds across CPU and CPU/XPU architectures.
• Enhance JIT compute graphs through operator fusion, memory layout optimizations, and tailored scheduling to minimize latency and maximize throughput.
• Preferred: Prior hands-on experience in LLVM or MLIR compiler development.

Model Profiling & Framework Acceleration

• Conduct end-to-end performance analysis across leading AI frameworks (e.g., PyTorch, TensorFlow, ONNX, and llama.cpp) to identify system bottlenecks.
• Design and execute high-impact acceleration strategies, including specialized kernel tuning and graph-level execution improvements.
• Preferred: Demonstrated success tuning AI models across diverse execution frameworks.

Technical Research & Insights

• Monitor and evaluate cutting-edge advancements in compiler design and AI infrastructure from academia and open-source communities.
• Deliver technical write-ups and benchmark reports outlining actionable optimization strategies and architectural trends.
• Preferred: Clear technical writing abilities with a history of research papers, blog posts, or analytical reports.

Key Requirements

• Strong proficiency in C/C++ alongside hands-on experience with compiler frameworks (LLVM, MLIR, or equivalent).
• Solid understanding of modern AI model topologies and real-time inference pipelines.
• Expertise using performance profiling toolchains (e.g., perf, VTune, TensorBoard).
• Knowledge of modern CPU/XPU architectures and hardware-level performance tuning.
• Sharp analytical thinking and pragmatic problem-solving capabilities.

Preferred Attributes

• BSc, MSc, or equivalent degree in Computer Science or a relevant STEM discipline.
• History of contributions to open-source compiler projects or communities (e.g., LLVM/MLIR).
• Familiarity with heterogeneous programming paradigms (CPU/GPU/XPU).
• Published articles, technical blogs, or papers focusing on AI compilers or systems optimization.
• Proactive self-starter with a passion for emerging tech and practical execution