Software Development Engineer, Amazon Q

Amazon Q is helping redefine how developers and cloud practitioners build, operate, troubleshoot, and optimize on AWS. We are building an intelligent assistant that goes beyond answering questions-combining generative AI with AWS context, retrieval, tools, and actions to help customers understand their environments and get work done faster.

Building these experiences creates challenging engineering problems. How do you ground an LLM"s response in relevant and up-to-date information from a customer"s AWS environment while maintaining strong security boundaries and low latency? How do you build agentic experiences that can reason, retrieve information, invoke tools, and take actions reliably and safely? How do you measure and improve the quality of AI-powered experiences when traditional software testing alone is not enough?

We are looking for a Software Development Engineer II to help build the next generation of Amazon Q. You will work at the intersection of generative AI, agentic systems, and large-scale distributed services, developing customer-facing capabilities used across AWS.

In this role, you will own meaningful features and components from design through production. You will work closely with other engineers, product managers, applied scientists, UX partners, and teams across AWS to translate customer needs into scalable technical solutions. You will contribute to architectural decisions, implement production-quality software, and continuously improve the reliability, performance, and quality of the systems you own.

This is an opportunity to work on emerging generative AI technologies while solving real distributed-systems problems at AWS scale.

Key job responsibilities

As a Software Development Engineer II on the Amazon Q team, you will:

Build: Design, implement, test, and launch customer-facing Amazon Q capabilities and the backend services that power them. Build systems that connect foundation models with AWS knowledge, customer context, retrieval systems, APIs, tools, and actions.

Solve: Work independently on complex engineering problems involving generative AI, distributed systems, contextual grounding, retrieval, tool execution, agentic workflows, latency, reliability, and security.

Own: Take ownership of features and services throughout their lifecycle-from requirements and technical design through implementation, deployment, operations, troubleshooting, and continuous improvement.

Design: Contribute to system architecture and technical designs, evaluate engineering trade-offs, participate in design reviews, and help build solutions that are scalable, maintainable, secure, and cost effective.

Operate: Build reliable production systems with strong engineering practices around testing, observability, monitoring, deployment, operational readiness, and incident response. Use operational and customer signals to identify and address opportunities for improvement.

Collaborate and innovate: Work closely with engineers, applied scientists, product managers, and partner teams to bring advances in foundation models, retrieval, reasoning, and agentic AI into production. Participate in code reviews, share technical knowledge, and help raise the engineering quality of the team.

A day in the life

You"ll start the day with the signals from how Amazon Q performed yesterday - satisfaction trends, latency tails, eval results - and pick up the problem that matters most. Mornings might be a design review with applied scientists on retrieval or agent orchestration; afternoons, writing the service code that ships it. You"ll partner with PMs, TPMs, and AWS service teams onboarding their own capabilities into the assistant, and you"ll own what you build in production. Expect to move between deep distributed-systems work, prompt and evaluation iteration, and A/B experiments that tell you whether your change actually helped a customer.

About the team

We"re the team behind Amazon Q in the AWS Management Console - engineers, applied scientists, and product managers working in one loop rather than in handoffs. Our mission is simple to say and hard to do: make the assistant genuinely useful to anyone operating on AWS. We ship fast, measure honestly, and say so when the data doesn"t support the idea we liked. Because the field moves monthly, we expect to rewrite our own assumptions often, and we"d rather learn from a real experiment than argue in a doc. We care about operational excellence, and about each other"s time.