Edge Gen AI Scientist - Quantization & Optimization

Amazon Inc.; Asti; Sur place; EUR 120 000 - 170 000; Plein temps

Résumé du poste

Amazon Devices is an inventive research and development company designing devices like Kindle, Fire tablets, Fire TV, Echo and Astro. This role focuses on developing edge Gen AI techniques and co-design with ML hardware to deliver state-of-the-art consumer products.

We seek exceptional scientists to join our Applied Science team, optimize edge models, and publish open-source work at leading conferences such as NeurIPS and MLSys.

Qualifications

• Bachelor's degree in computer science, mathematics, statistics, machine learning or equivalent quantitative field.
• Experience programming in Java, C++, Python or related language.

Responsabilités

• Quantize, prune, distill, finetune Gen AI models to optimize for edge platforms.
• Fundamentally understand Amazon’s underlying Neural Edge Engine to invent optimization techniques.
• Analyze deep learning workloads and provide guidance to map them to Amazon’s Neural Edge Engine.
• Use first principles of Information Theory, Scientific Computing, Deep Learning Theory, Non Equilibrium Thermodynamics.
• Train custom Gen AI models that beat SOTA and paves path for developing production models.
• Collaborate closely with compiler engineers, fellow Applied Scientists, Hardware Architects and product teams to build the best ML-centric solutions for our devices.
• Publish in open source and present on Amazon's behalf at key ML conferences - NeurIPS, ICLR, MLSys.

Connaissances

Java

C++

Python

Formation

Bachelor's degree in computer science, mathematics, statistics, machine learning or equivalent quantitative field

Outils

Java

C++

Python

Description du poste

Amazon Devices is an inventive research and development company designing devices like Kindle, Fire tablets, Fire TV, Echo and Astro. This role focuses on developing edge Gen AI techniques and co-design with ML hardware to deliver state-of-the-art consumer products.

We seek exceptional scientists to join our Applied Science team, optimize edge models, and publish open-source work at leading conferences such as NeurIPS and MLSys.