Applied Scientist: Hardware-Aware ML for ML Accelerators

Amazon Web Services (AWS); Cupertino (CA); On-site; USD 133,000 - 222,000; Full time

Job summary

Annapurna Labs (U.S.) Inc. within Amazon AWS Neuron is seeking Applied Scientists to advance hardware-aware ML for Trainium and Inferentia.

You own model and algorithm decisions end-to-end from research through production, publish at top venues, and mentor engineers in a strategic growth area for AWS. You will develop production-quality code in PyTorch or JAX, integrate components into large-scale training and inference systems, and collaborate with foundation-model, engineering, and hardware

Qualifications

• PhD or Master’s with 4+ years in CS/CE/ML or related field.
• Experience building models for business applications.
• Proficiency in Java, C++, Python.
• Experience with state-of-the-art DL architectures, training and optimization, and model pruning.
• Publications or patents in top-tier venues or journals.

Responsibilities

• Own scientific problems end-to-end—from research to production impact.
• Develop production-quality code in PyTorch or JAX and integrate components into large-scale training/inference systems.
• Partner with foundation-model, engineering, and hardware teams to inform Trainium/product stack.
• Mentor scientists and interns, perform peer reviews, shape team goals and roadmap.
• Publish research at top venues (ICLR, NeurIPS, ICML, MLSys).

Skills

Deep learning model design

ML training optimization

Distributed computing

Education

PhD, or Master’s degree with 4+ years in CS/CE/ML or related field

Tools

Java

C++

Python

Job description

Annapurna Labs (U.S.) Inc. within Amazon AWS Neuron is seeking Applied Scientists to advance hardware-aware ML for Trainium and Inferentia.

You own model and algorithm decisions end-to-end from research through production, publish at top venues, and mentor engineers in a strategic growth area for AWS. You will develop production-quality code in PyTorch or JAX, integrate components into large-scale training and inference systems, and collaborate with foundation-model, engineering, and hardware