Edge Gen AI Scientist: Model Optimization & Deployment

Amazon; Hinoba-an; On-site; PHP 900,000 - 1,500,000; Full time

Job summary

Amazon Devices is seeking exceptional scientists to join the Applied Science team, focusing on Gen AI on edge for consumer devices and co-design with ML HW. You will work on quantization, pruning, distillation, and training of production-grade models for edge platforms.

The role requires strong ML background, PhD or MSc with research, and experience publishing or patenting work. Join a team that publishes and presents at top conferences and collaborates across engineering and product groups.

Qualifications

• 2+ years building ML models for business use.
• PhD or Master plus 2+ years applied research.
• Publications or patents in top venues preferred.
• Proficiency in Java, C++, Python or similar.
• Strong math background: linear algebra, calculus, statistics.

Responsibilities

• Quantize, prune, distill, and finetune Gen AI models for edge devices.
• Understand and optimize Neural Edge Engine concepts.
• Map deep learning workloads to edge hardware and software.
• Apply information theory, scientific computing, and DL theory to models.
• Train Gen AI models that beat SOTA and aid production models.
• Collaborate with compiler engineers, hardware architects, and product teams.
• Publish open-source work and present at ML conferences (NeurIPS, ICLR, MLSys).

Skills

ML programming

Java/C++/Python

Mathematics

Publications/patents

Algorithms & data structures

Education

PhD or MSc + 2+ years research

Tools

None

Job description

Amazon Devices is seeking exceptional scientists to join the Applied Science team, focusing on Gen AI on edge for consumer devices and co-design with ML HW. You will work on quantization, pruning, distillation, and training of production-grade models for edge platforms.

The role requires strong ML background, PhD or MSc with research, and experience publishing or patenting work. Join a team that publishes and presents at top conferences and collaborates across engineering and product groups.