Senior Quantization Engineer - Edge AI Model Optimization

Job Title

Senior Quantization Engineer - Edge AI Model Optimization

Role Summary

Work on model quantization and complementary optimization techniques to enable efficient on-device inference for NXP's Ara2 NPUs. The role spans research, prototyping, and production deployment for CNNs, LLMs and VLMs.

Collaborate with AI research, hardware engineering and software teams to convert research methods into robust, low-latency, memory- and compute-efficient implementations.

Experience Level

Senior. The posting targets experienced engineers/researchers; no explicit years-of-experience requirement stated.

Responsibilities

Key responsibilities include research, prototyping, productionization, cross-functional integration, and IP generation.

• Survey and evaluate recent model optimization research (e.g., NeurIPS, ICLR, CVPR) with emphasis on quantization and compression techniques.
• Prototype and adapt state-of-the-art methods to NXP hardware constraints and demonstrate proof-of-concept results.
• Translate prototypes to production-grade, optimized code in C++/Python, meeting strict memory and compute efficiency targets.
• Document algorithmic tradeoffs, create deployment recipes, and mentor engineers on numerical methods and optimization.
• Serve as the technical liaison between AI research and hardware teams, providing quantitative guidance on accuracy vs. performance tradeoffs.
• Contribute to intellectual property through patents and technical publications.

Requirements

Must-have technical skills and domain experience; preferred items listed as nice-to-have.

• Must-have: Practical AI/ML experience with deep learning models including CNNs and transformer-based generative models.
• Must-have: Hands-on experience with PyTorch, ONNX, and model conversion/optimization pipelines.
• Must-have: Strong software engineering skills in Python and C++, and familiarity with production development practices.
• Must-have: Understanding of embedded system constraints (latency, power, memory bandwidth) and how models map to hardware.
• Nice-to-have: Experience with advanced quantization methods (e.g., GPTQ, SpinQuant) and quantization for generative models.
• Nice-to-have: Experience with NPUs, device-level profiling, diagnosing memory bottlenecks, or custom kernel development.
• Nice-to-have: Knowledge of compiler/runtime technology such as MLIR or TVM.

Education Requirements

MSc required or PhD preferred in Computer Science, Electrical Engineering, Mathematics, or a related field with emphasis on Machine Learning / Deep Learning.

About the Company

Company: NXP Semiconductors

Headquarters: Nijmegen, Netherlands

NXP Semiconductors N.V. is a global semiconductor company that provides High Performance Mixed Signal and Standard Product solutions. With over 45,000 employees and operations in more than 35 countries, NXP is a leader in secure connectivity solutions for embedded applications, catering to automotive, industrial IoT, mobile, and communication infrastructure markets. The company is committed to innovation and sustainability, advancing a smarter, safer, and more sustainable world through technology.