AWS Documentation, one of the most visited technical documentation sites in the world, is increasingly authored using AI agents; not as an experiment, but in production. You'll be the senior engineer owning the architecture behind it: a runtime that executes documentation workflows autonomously, multi-agent orchestrated authoring tools that help service teams produce documentation from their design documents and service models, and analyzers that review those contents against AWS standards before commit.
This is hard in interesting ways: agent that writes plausible-sounding API documentation is worse than no agent at all, so correctness and verifiability sit at the heart of the architecture. You'll decide how specialized agents divide work versus deterministic orchestration with agentic reasoning at the leaves, how their output gets evaluated, and when a workflow needs to escalate to a human. In a domain where agent patterns change every few months, you know that today's right call has a short shelf life and adapt your own orchestration patterns, tool ecosystem design, and inference cost and latency.
Shape technical strategy where ambiguity spans both product shape and technical approach, lead initiatives through engineers and teams you don't control, and ship weekly while doing it.
Key job responsibilities
- Own the architecture across the portfolio — hosted agent runtime, multi-agent authoring workflows, commit-time content analyzers, and operational telemetry — making technology choices where neither the product shape nor the technical approach is defined
- Design the correctness and verifiability layer: ground agent output in authoritative sources, catch fabrication, and define when a workflow escalates to human review
- Lead projects requiring work from multiple engineers, including agent orchestration patterns, evaluation frameworks, bidirectional XML transformation, and developer-facing authoring tools
- Turn content quality evaluations into feedback signals that measurably improve agent behavior, and influence architectural decisions in adjacent systems your services depend on but don't control
- Raise the engineering bar through design reviews, exemplary code, and mentoring engineers newer to production AI systems, while identifying root causes of systemic issues and balancing delivery speed against long-term system quality
A day in the life
You'll start with yesterday's agent runs - did the research agent find the right sources, did the writing agent produce content that passes style review? You're asking whether the orchestration architecture still holds as model capabilities shift.
Mid-morning, a service team needs documentation generated from a model type your pipeline doesn't handle. Neither the product shape nor the approach is obvious, so you design the pattern and prototype it.
After lunch, a partner team's API redesign affects how your agents integrate. You influence that design to serve both human and agentic workflows - building consensus without owning it.
About the team
The Applied Science team builds the AI-powered tools that are fundamentally changing how AWS documentation gets created. We develop and operate multi-agent content generation systems, intelligent code-review analyzers, developer-facing authoring tools, and the telemetry infrastructure that measures it all.
Our tools are used daily by hundreds of contributors across AWS. We move fast — shipping improvements weekly — while maintaining the reliability that production content systems require. The team values pragmatic engineering, measurable impact, and the intellectual honesty to know when an AI approach isn't working and to try something else.
We're part of TCX Engineering, the organization behind the AWS Documentation website and the AI-first content platform powering it. You'll work closely with product managers, content strategists, and peer engineering teams building the broader platform infrastructure.
