AI Compiler Engineer

Job Title: AI Compiler Engineer Job Description: Overview: We are seeking a skilled and motivated AI Compiler Engineer to lead the development of our AI compiler for the Neural Processing Unit (NPU). As part of our team, you will play a critical role in transforming high-level neural network workloads (such as CNNs and transformer models) into efficient machine code that maximizes inference performance, minimizes power consumption, and optimizes memory usage. Responsibilities: 1. Compiler Development: • Design and develop an AI compiler that translates high-level neural network descriptions (in frameworks like TensorFlow, PyTorch, or ONNX) into optimized machine code for the NPU. • Implement parsing algorithms to analyze neural network graphs and extract relevant information for compilation. 2. Optimization Strategies: • Optimize the computational graph by applying techniques such as operator fusion, kernel selection, and memory layout transformations. • Minimize inference time, reduce power consumption, and optimize memory footprint through intelligent code generation. 3. Performance Profiling and Analysis: • Profile compiled code to identify bottlenecks and areas for improvement. • Collaborate with hardware architects to understand NPU microarchitecture and tailor optimizations accordingly. 4. Code Generation and Lowering: • Generate efficient machine code from the intermediate representation of neural network operations. • Ensure compatibility with the NPU's instruction set architecture. 5. Integration and Testing: • Integrate the AI compiler into the overall software stack, including runtime libraries and drivers. • Develop test suites to validate correctness, performance, and compatibility across various neural network models. 6. Documentation and Communication: • Document design decisions, algorithms, and implementation details. • Collaborate with cross-functional teams, including hardware engineers, software developers, and researchers. Qualifications: • Master's or Ph.D. in Computer Science, Electrical Engineering, or a related field. • Minimum of 3 years of experience in compiler development, preferably with exposure to AI or machine learning. • Proficiency in C/C++/Python programming and familiarity with compiler construction. • Knowledge of neural network frameworks (TensorFlow, PyTorch, etc.) and their intermediate representations. • Experience with performance profiling tools and optimization techniques. • Understanding of machine learning concepts and hardware architectures. • Familiarity with data structures, optimization algorithms, and DevOps tools (such as Git, CI/CD pipelines). • Strong problem-solving abilities and attention to detail. • Excellent communication skills for collaborating with cross-functional teams. Preferred Skills: • Familiarity with deep learning frameworks and libraries. • Knowledge of hardware acceleration technologies and frameworks. • Experience of tensor computation and optimization. • Previous work in compiler for CPU, GPU, DSP, NPU is a plus. If you are passionate about compiler engineering, AI optimization, and want to be part of a team driving innovation, we encourage you to apply. Join us in shaping the future of AI acceleration.