ML Compiler Engineer, Annapurna Labs

Overview

The AWS Neuron Compiler team is actively seeking skilled compiler engineers to join our efforts in developing a state‑of‑the‑art deep learning compiler stack. This stack optimizes models from PyTorch, Tensor Flow, JAX across domains such as Large Language Models and Vision. Your role involves working closely with our custom Machine Learning accelerators (Inferentia/Trainium) and contributing to system‑wide performance optimization.

• Design, develop, and optimize compiler features for deep learning models.
• Tackle technical challenges across the stack, from instruction scheduling and memory allocation to graph partitioning and code generation.
• Collaborate with Runtime, Frameworks, and Hardware teams to deliver end‑to‑end performance improvements.
• Research and implement solutions that provide the best experience for customers.
• Participate in design discussions, code review, and cross‑functional decision making.

• Bachelor's degree in Computer Science, Engineering, Mathematics, or related field.
• 3+ years of professional software development experience.
• Proficiency in C++ (or other programming languages).
• 2+ years of experience with optimization algorithms, graph theory, hardware bring‑up, FPGA placement/routing, or hardware resource management.

• 3+ years of experience in the full software development life cycle.
• M.S. or Ph.D. in computer science or related field.
• Strong knowledge in compiler design, instruction scheduling, memory allocation, data‑transfer optimization, graph partitioning, parallel programming, code generation, ISA design, hardware bring‑up, or hardware‑software co‑design.
• Experience with LLVM and/or MLIR.
• Experience developing algorithms for simulation tools.
• Experience with Tensor Flow, PyTorch, and/or JAX.
• Experience with large‑language models, vision, or other deep‑learning models.

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