About the job
As the SDM for the Neuron Inference Technology building blocks team, you will guide your expert AI engineers to build fundamental inference technology building blocks and libraries to enable AI developers to optimize model for inference on Trainium and Inferentia devices. We’re currently focusing on MoE models such as GPT OSS for Trainium 2 and the upcoming Trainium 3. You will develop and optimize blocks such as attention kernels and deliver them in the Neuronx_Distributed Inference Libraries, enabling customers to optimize LLMs, multimodal, and generative models.
Responsibilities
You will work with your senior management and technical leaders to define the building blocks for the latest LLMs, build and deliver them to customers. You will manage changing priorities as new models and new technologies emerge, and you adapt your team’s work to manage them. You will dive deep to help your team solve technical challenges.
Qualifications
Minimum
Established background in optimizing LLMs, such as delivering high-performance models using distributed inference libraries; capable of managing demanding, fast-changing priorities; strong technical ability to understand and deliver as part of a vertically integrated system stack consisting of the PyTorch inference library, Neuron compiler, runtime and collectives.
Preferred
The ideal candidate will have an established background in optimizing LLMs, such as delivering high-performance models using distributed inference libraries. You should be capable of managing demanding, fast-changing priorities. You should have a strong technical ability to understand and deliver as part of a vertically integrated system stack consisting of the PyTorch inference library, Neuron compiler, runtime and collectives.