Software Development Engineer III, Machine Learning Annapurna - System Integration

Do you want to build AI-powered tooling that keeps custom silicon running reliably at scale? In Annapurna Labs at Amazon Web Services (AWS), we sit at the intersection of software and hardware, designing the systems that detect failures, generate repair instructions, and automate qualification for Trainium products. As a Software Development Engineer III, you will lead the technical design and delivery of this tooling platform - combining generative AI techniques with deep hardware knowledge to solve problems that few teams in the industry encounter. With hardware complexity and scope expanding rapidly, this is a rare opportunity to shape how we bring the next generation of custom chips to operational stability.

Key job responsibilities

• Design and deliver generative AI-based tools that identify complex hardware failures, dynamically create repair instructions from hardware system designs, and automate the qualification framework for post-new-product-introduction hardware changes on Trainium systems.
• Own system design decisions across the tooling software stack, evaluating tradeoffs in scalability, maintainability, and performance while partnering with silicon design and operations teams on cross-functional technical discussions.
• Collaborate with hardware design and sustaining teams to align tooling updates with upcoming silicon changes, creating the right integration checkpoints to ensure successful delivery.
• Drive adoption of engineering best practices across the team through code reviews, design reviews, and mentorship - raising the quality bar and helping fellow engineers grow.
• Evaluate and refine generative AI model outputs used in the tooling platform, applying responsible AI practices to ensure accuracy and reliability at production scale.

A day in the life

You start your morning reviewing system health metrics for the tooling platform, then join a design review where a teammate walks through a new generative AI model for automated failure diagnosis on Trainium hardware. You ask probing questions that sharpen the approach. After lunch, you pair with an engineer to work through a tricky data-modeling problem, then shift to a cross-team sync with hardware design engineers to align on upcoming silicon changes that will require updates to your qualification automation. Later, you write a technical document outlining how the tooling will scale to support the next generation of Trainium products.

About the team

Our team within Annapurna Labs builds the AI-powered software tooling that ensures Trainium systems meet operational stability targets before handoff to sustaining teams. We work at the intersection of software and silicon alongside systems development engineers, software development engineers, and hardware engineers - a cross-functional mix that is rare even within AWS. As Annapurna Labs continues to expand its custom chip portfolio, we are investing in new generative AI capabilities and scaling our platform to keep pace. If you are excited about solving problems where software meets hardware at industry-leading scale, we would love to hear from you.