Join the Amazon AGI team to revolutionize the future of AI by developing breakthrough technologies that dramatically improve the efficiency of large language and foundation models. Work at the frontier of AI research, designing novel approaches for model compression, efficient architectures, long context extension and training optimization that will shape the next generation of AI systems.
We're seeking exceptional Applied Scientists to join our team dedicated to solving one of AI's most critical challenges: making foundation models more efficient through lower latency and higher throughput. You'll work on research and new technology development while collaborating with world-class researchers and engineers.
- Pioneer new approaches to foundation models
- Publish and present research at top-tier conferences and journals
- Work with state-of-the-art LLMs and multi-modal foundation models
- Access to substantial computational resources for research
- Collaborate across AGI teams to implement solutions at scale
Key job responsibilities
- Research and develop novel techniques for efficient runtime inference (low latency, high throughput)
- Design and evaluate efficient foundation model architectures
- Create new methods for improving training efficiency
- Conduct experimental studies to validate efficiency improvements
- Write high-quality Python code to implement research ideas
- Collaborate with team members to integrate solutions into production systems
- Author technical documentation and research papers
- Present findings to technical and non-technical stakeholders
A day in the life
Your day might start with a team stand-up to discuss ongoing projects and brainstorm solutions to technical challenges. You'll spend time implementing and testing new efficiency optimization techniques in Python, analyzing performance metrics, and iterating on approaches. You'll collaborate with team members to review code and research results, participate in technical discussions about architecture designs, and engage with other AGI teams to understand their efficiency needs. You might end your day analyzing experimental results or writing up findings for a research paper. Throughout the week, you'll have opportunities to present your work to stakeholders and contribute to the team's research roadmap.
Our stakeholders include other AGI teams within Amazon, and our work directly impacts the cost and performance of AI systems used across Amazon's products and services. You'll be solving complex technical challenges that make advanced AI more accessible and efficient at scale.
About the team
The Neural Efficiency Science team was founded in 2024 to develop new technologies that improve the cost-to-performance ratio of AGI foundation models. Amazon believes in customer-obsessed research and as a large number of AI applications are getting built, we see a need to invest in making the underlying foundation models serving these applications cheaper, faster and more accurate at a given price point. Our team keeps up with all of the latest research and trends in what we call "neural efficiency" and publishes our own findings at top conferences such as ICML and NeurIPS while also making our technology available in production for internal and external customers.
We're seeking exceptional Applied Scientists to join our team dedicated to solving one of AI's most critical challenges: making foundation models more efficient through lower latency and higher throughput. You'll work on research and new technology development while collaborating with world-class researchers and engineers.
- Pioneer new approaches to foundation models
- Publish and present research at top-tier conferences and journals
- Work with state-of-the-art LLMs and multi-modal foundation models
- Access to substantial computational resources for research
- Collaborate across AGI teams to implement solutions at scale
Key job responsibilities
- Research and develop novel techniques for efficient runtime inference (low latency, high throughput)
- Design and evaluate efficient foundation model architectures
- Create new methods for improving training efficiency
- Conduct experimental studies to validate efficiency improvements
- Write high-quality Python code to implement research ideas
- Collaborate with team members to integrate solutions into production systems
- Author technical documentation and research papers
- Present findings to technical and non-technical stakeholders
A day in the life
Your day might start with a team stand-up to discuss ongoing projects and brainstorm solutions to technical challenges. You'll spend time implementing and testing new efficiency optimization techniques in Python, analyzing performance metrics, and iterating on approaches. You'll collaborate with team members to review code and research results, participate in technical discussions about architecture designs, and engage with other AGI teams to understand their efficiency needs. You might end your day analyzing experimental results or writing up findings for a research paper. Throughout the week, you'll have opportunities to present your work to stakeholders and contribute to the team's research roadmap.
Our stakeholders include other AGI teams within Amazon, and our work directly impacts the cost and performance of AI systems used across Amazon's products and services. You'll be solving complex technical challenges that make advanced AI more accessible and efficient at scale.
About the team
The Neural Efficiency Science team was founded in 2024 to develop new technologies that improve the cost-to-performance ratio of AGI foundation models. Amazon believes in customer-obsessed research and as a large number of AI applications are getting built, we see a need to invest in making the underlying foundation models serving these applications cheaper, faster and more accurate at a given price point. Our team keeps up with all of the latest research and trends in what we call "neural efficiency" and publishes our own findings at top conferences such as ICML and NeurIPS while also making our technology available in production for internal and external customers.