Principal Applied Scientist, ML Codesign
Core
Define joint optimization of model compression and silicon architecture for Amazon's next-generation edge and cloud inference accelerators.
Role type
Principal Applied Scientist (ML Codesign)
Builds
Next-generation ASICs and compression algorithms for multi-billion parameter language models
Domain
Machine Learning Systems / Computer Architecture / Semiconductor Design
Deliverable
production ML models
Required skills
hardware-aware neural architecture search, low-bit quantization, structured pruning, knowledge distillation, hardware-aware training, computer architecture fundamentals, silicon architecture definition, large-scale model optimization
Preferred skills
hardware-software codesign, Mixture-of-Experts inference, RTL review, MLIR/OpenXLA familiarity, vertically integrated stack experience
Technologies
ASIC, FPGA, MLIR, OpenXLA
Responsibilities
Define hardware-aware compression roadmap, own joint optimization of compression algorithms with hardware, represent applied science in silicon architecture reviews, set science roadmap for compression techniques, mentor senior and mid-level applied scientists, serve as single-threaded technical leader for codesign agenda
Seniority
Principal, hands-on IC with strategy & mentorship
