Job Title:
Quantizer, Software Engineer
Company:
NXP Semiconductors México, S. de R.L. de C.V.
We at NXP have an environment that fosters innovation. Our team has technology experts who understand the big picture and mentors who coach passionate professionals to work on the most exciting challenges. We share responsibilities in everything we do, where every point of view is valued. Join us!
Job Summary
Our game-changing AI solutions revolutionize what people and businesses can achieve. NXP inference processorscombined with our SDK deliver unrivalled deep learning performance at the edge to accelerate and optimizereal-time decision making where every millisecond is critical, and power efficiency is a must. NXP solutionsembed high-performance AI into edge devices to create a smarter, safer, and more enjoyable world. Edge AI is onthe brink of a boom, and NXP is looking forward to playing a significant role in it.We’re hiring a hands-on Quantizer Engineer with 4–8 years of experience to own and evolve production quantization infrastructure for our silicon.
You will design and implement advanced PTQ/QAT, calibration and mixed-precision algorithms, add features to our C++ quantizer, and drive integration with compilers, runtimes and kernels so real customer models run efficiently and reliably on our hardware.Now tell us your story. We are looking forward to reviewing your application.Make your mark!
Job Responsibilities
- Design and implement quantization features and passes in the production quantizer (C++), including per-tensor / per-channel quantization, asymmetric/symmetric schemes, and mixed-precision flows.
- Build and maintain PTQ and QAT pipelines, calibration strategies, and representative-dataset samplers that produce reproducible accuracy/perf tradeoffs.
- Develop approximation & numerical algorithms (range estimation, bias-correction, BN-folding) with rigorous mathematical justification.
- Integrate quantization metadata and ensure correct execution on target hardware (accumulator widths, saturation, rounding behavior).
- Implement robust unit tests, deterministic calibration, fuzz tests for edge-cases, and CI hooks for accuracy/regression monitoring.
- Profile and optimize C++ code paths to meet memory- and compute-constrained edge targets.
- Work cross-functionally with compiler, kernel, firmware and HW architects on software–hardware co-design tradeoffs; provide quantified guidance on ISA/ABI and micro-architectural impacts.
- Run POCs and prototype novel quantization ideas, evaluate results, and turn successful experiments into deployable recipes and tooling.
- Document quantization recipes, tuning knobs, and deployment handoffs; mentor junior quantizers and share best practices.
Job Qualifications
- M.S, M.Eng or Ph.D in Computer Science, Electrical Engineering, or Mathematics with a focus on Machine Learning, Deep Learning, or Numerical Analysis
- 4–8 years of professional experience in ML systems, inference engineering, compilers, or related software engineering roles with demonstrable quantization experience.
- Strong C++ (production) skills and solid Python for data/experimentation tooling.
- Deep mathematical foundation: linear algebra, probability/statistics, numerical methods, and approximation algorithms.
- Hands-on experience implementing PTQ and QAT workflows and measuring/troubleshooting accuracy regressions on real models (transformers, convnets, etc.).
- Practical knowledge of model formats and toolchains (ONNX, TorchScript, TensorFlow) and integration with compilers/runtimes.
- Familiar with low-level quantization constraints: fixed-point arithmetic, accumulator sizing, per-channel scales, symmetric/asymmetric zero-points, and rounding/saturation behaviors.
- Strong debugging, profiling and performance-analysis skills; experience with unit testing frameworks and CI integration.
- Excellent written and verbal communication: able to explain algorithmic trade-offs and produce actionable design docs.
Preferred Qualifications
- Experience with MLIR/TVM/Glow/LLVM or other compiler frameworks and embedding quantization metadata in IR.
- Prior work on learned quantization methods (QAT variants), per-channel learned scales, or calibration automation.
- Exposure to embedded systems, firmware, and runtime/kernel-level integration on NPUs/accelerators.
- Contributions to open-source quantization or compiler projects, or relevant publications.
- C++ experience in memory-constrained or real-time environments
What You Will Gain
- Be part of a pioneering team shaping the future of AI and edge computing.
- Work on innovative projects that solve real-world challenges.
- Opportunity to grow with a dynamic, forward-thinking company.
- Competitive salary, benefits, and a collaborative work environment.
#LI-FCC3More information about NXP in Mexico...#LI-fcc3
