Design, implement, and optimize compiler features for Intel NPU architectures, ensuring high performance and efficient code generation.
Enable and optimize machine learning models—including LLMs—for deployment on Intel NPUs, focusing on quantization, dynamic execution, and hardware-specific acceleration.
Collaborate with other teams to productionize new models and support emerging AI workloads.
Develop and maintain model conversion, quantization, and deployment pipelines, ensuring correctness, reproducibility, and compliance with Intel’s standards.
Benchmark, profile, and debug models and software components to identify and resolve performance bottlenecks.
Stay current with the latest advancements in compilers, model optimization, quantization, and LLM research, and integrate best practices into the NPU software stack.
Write clear technical documentation and provide support to internal and external users.
Qualifications:
Bachelor’s, Master’s, or Ph.D. in Computer Science, Electrical Engineering , or related field.
Proven expertise in C/C++ with strong software design and optimization skills.
Solid understanding of AI model optimization techniques such as quantization, pruning, and distillation.
Familiarity with large language models and their deployment requirements.
Knowledge of computer architecture , hardware acceleration, and low-level performance tuning .
Proficiency with Linux environments , virtualization, and CI/CD workflows.
Strong analytical, problem-solving, and cross-team collaboration skills in fast-moving technical settings.
Experience with modern compiler infrastructures (e.g., LLVM, MLIR ) or code generation for custom accelerators is a plus.
Hands-on experience with AI frameworks (OpenVINO, TensorFlow, PyTorch, ONNX), Python , and performance tools for NPUs, GPUs, or FPGAs is a plus.
Experience developing dynamic execution or runtime systems that handle variable input sizes and adaptive behavior is a plus.
Familiarity with collaborative tools (GitHub, Jira) and open-source contribution practices is a plus.