Machine Learning Engineer (Quantization)

Machine Learning Engineer (Quantization)

Dresden, Germany (on-site)

IC Resources is seeking a Machine Learning Engineer with expertise in model quantization for hardware, to join our client's innovative team in Dresden, Germany. This role offers a unique opportunity to work at the intersection of machine learning and hardware design, shaping the future of AI hardware for edge devices. The successful candidate will play a crucial role in developing a scalable inference framework and collaborating on cutting-edge quantization techniques.

Primary Responsibilities:

• Design and maintain a highly optimised inference framework tailored to cutting-edge AI hardware.
• Collaborate closely with ML, compiler, and hardware teams to refine and implement advanced quantization algorithms.
• Innovate on state-of-the-art quantization methods such as AdaRound, BRECQ, GPTQ, and QuaRot, adapting these techniques to meet unique hardware requirements.

Essential Experience:

• Proficiency in PyTorch, including experience with torch.FX.
• Ability to develop efficient custom CUDA kernels.
• Strong understanding of current research in quantization techniques and practical experience applying these methods.

Desired Experience:

• Familiarity with neural network compression methods like Adaround, QDrop, QUIP, or GPTQ.
• Knowledge of ML tools such as Hugging Face Transformers or DeepSpeed.

What’s On Offer:

• Be part of a forward-thinking team driving innovation in AI hardware.
• Opportunity to contribute to fundamental architectural decisions and open-source projects.
• A dynamic, collaborative environment at the cutting edge of machine learning and hardware integration.

How to Apply:
If you’re excited by the opportunity to advance your career as a Machine Learning Engineer focussed on quantization, and contribute to ground-breaking technology, apply now for immediate consideration. Contact Chris Wyatt, Principal Recruitment Consultant, for more information and assistance with your application.