Software Development Engineer, AI Core Infra - AI Analytics and Insights

DESCRIPTION

Amazon Advertising is one of Amazon's fastest growing and most profitable businesses, responsible for defining and delivering a collection of advertising solutions that drive product discovery and sales. We deliver billions of ad impressions every single day on behalf of our advertisers and generate massive volumes of data across the advertising lifecycle.

We are building a conversational AI assistant that empowers data engineers, analysts, and business stakeholders to interact with our vast advertising data lake through natural language. Using state-of-the-art generative AI, SpektrBot enables teams to generate insights faster by translating natural language queries into accurate, trusted data retrievals, eliminating the friction between questions and answers.

As an SDE II on this team, you will be at the center of designing, building, and scaling the systems that power this conversational AI experience. You will architect robust backend services, build reliable integrations with large language models (LLMs), develop retrieval-augmented generation (RAG) pipelines, and engineer the feedback loops that allow our system to continuously evaluate and improve itself. You will work alongside applied scientists, product managers, and fellow engineers in a fast-paced, high-ownership environment where your contributions directly shape the product and influence how thousands of internal customers interact with data.

This is a rare opportunity to build a 0-to-1 product at Amazon's scale, one that sits at the intersection of generative AI, data engineering, and conversational interfaces. If you are passionate about building production-grade AI-powered systems, thrive in ambiguity, and want to see your work used every day by teams across Amazon Advertising, we'd love to talk to you.

At Amazon an SDE can expect to design flexible and scalable solutions, and work on some of the most complex challenges in large-scale computing by utilizing your skills in data structures, algorithms, and object oriented programming. Coming to Amazon gives you the opportunity to work on a small development team in one of our many organizations; Amazon Web Services, ecommerce Services, Kindle, Marketplace, Operations, Platform Technologies and Retail.

Key job responsibilities

• Design and build scalable services that power SpektrBot's conversational AI platform, including query orchestration, context management, session handling, and response generation pipelines
• Engineer RAG architectures — build and optimize retrievers, vector stores, embedding pipelines, and metadata indexing systems to ensure accurate and contextually relevant data retrieval from our advertising data lake
• Integrate and operationalize LLMs — build the infrastructure for prompt management, model routing, multi-agent orchestration, and chain-of-thought workflows that enable natural, multi-turn conversations
• Build automated evaluation and feedback loops — design systems that continuously measure response accuracy, detect regressions, and feed corrections back into the system to drive improvement over time
• Develop SQL generation and validation pipelines — engineer the end-to-end flow from natural language intent to generated SQL, including guardrails, query validation, and result verification to ensure trusted outputs
• Build tooling for metadata auto-curation — create LLM-powered tools that automatically enrich, classify, and maintain the metadata catalog that underpins accurate data retrieval
• Collaborate with applied scientists to productionize NLP/NLU models and integrate them into the bot's overall architecture
• Own systems end-to-end — from design through deployment, monitoring, alarming, and operational excellence in a production environment serving internal customers at scale

A day in the life

You might start by reviewing your automated evaluation dashboard, investigating a flagged drop in SQL generation accuracy and shipping a fix. After lunch, you pair with an applied scientist to integrate a new intent classification model into the orchestration layer, ensuring graceful fallback behavior. Later, you lead a design review for a multi-agent architecture that breaks complex user questions into parallel sub-queries. You end the day improving prompt templates based on user feedback patterns. Every day brings a different problem, but the thread is always building AI-powered systems that are accurate, reliable, and delightful to use.

About the team

You will invent new experiences and influence customer-facing shopping experiences to help suppliers grow their retail business and the auction dynamics that leverage native advertising; this is your opportunity to work within the fastest-growing businesses across all of Amazon! Define a long-term science vision for our advertising business, driven from our customers' needs, translating that direction into specific plans for research and applied scientists, as well as engineering and product teams. Th