Machine Learning - Compiler Engineer , AWS Neuron, Annapurna Labs

Role Overview

We are building AWS Neuron, an SDK that optimizes complex machine‑learning models for AWS Inferentia and Trainium chips. This role is for a software engineer on the Compiler team who will design and implement the next‑generation Neuron compiler, transforming models written in frameworks such as PyTorch, Tensor Flow, and JAX into efficient code for our custom accelerators.

• Design, implement, test, deploy and maintain the Neuron compiler to improve performance, stability, and usability.
• Partner with chip architects, runtime/OS engineers, scientists, and ML application teams to deliver optimized ML workloads on AWS accelerators.
• Develop compiler optimization and verification passes that support large‑scale models, including LLMs, vision transformers, and stable diffusion.
• Integrate open‑source tools such as OpenXLA, StableHLO, and MLIR to advance compiler capabilities.
• Build tools for numerical‑error analysis, defect resolution, and performance profiling.

The role involves designing compiler passes, implementing features for developers, analyzing numerical errors, troubleshooting compiler defects, and coordinating across multiple teams in a fast‑paced environment.

• 3+ years of professional software development experience.
• 2+ years of system design or architecture experience.
• Proficiency in at least one object‑oriented language such as C++ or Java.

• Master’s degree or PhD in Computer Science or a related field.
• 3+ years of production‑grade coding in C++/Java.
• Experience in compiler design for CPUs, GPUs, or ML accelerators.
• Familiarity with LLVM/MLIR and open‑source compiler tool chains.
• Experience with PyTorch, OpenXLA, StableHLO, JAX, TVM, and deep-learning model development.
• Experience with modern build systems such as Bazel or CMake.

Amazon is an equal‑opportunity employer and does not discriminate on the basis of protected veteran status, disability, or any other legally protected status.