Sr Data Scientist , Annapurna Labs, ML Acceleration Power Architecture

Amazon

Amazon

Software Engineering, IT, Data Science
Cupertino, CA, USA
Posted on Oct 30, 2025

Description

We are seeking a highly skilled Data Scientist to join our Machine Learning Architecture team, focusing on power and performance optimization for ML acceleration workloads across Amazon's global data center infrastructure. This role combines advanced data science techniques with deep technical understanding of ML hardware acceleration to drive efficiency improvements in training and inference workloads at massive scale.

Key job responsibilities
ata Analysis & Optimization

* Analyze power consumption and performance metrics across all Amazon data centers for machine learning acceleration workloads
* Develop predictive models and statistical frameworks to identify optimization opportunities and performance bottlenecks
* Create automated monitoring and alerting systems for power and performance anomalies

Strategic Planning & Deployment Guidance

* Provide data-driven recommendations for server deployments and capacity planning decisions across Amazon's global data center network
* Develop optimization scenarios and business cases to improve capacity delivery efficiency to customers worldwide
* Support strategic decision-making through comprehensive analysis of power, performance, and cost trade-offs

Cross-Functional Collaboration

* Partner with software engineering teams to optimize ML frameworks, drivers, and runtime systems
* Collaborate with hardware engineering teams to influence chip design, server architecture, and cooling system optimization
* Work closely with data center operations teams to implement and validate optimization strategies

Research & Development

* Conduct applied research on emerging ML acceleration technologies and their power/performance characteristics
* Develop novel methodologies for measuring and improving energy efficiency in large-scale ML workloads
* Publish findings and contribute to industry best practices in sustainable ML infrastructure