Amazon Bedrock AgentCore is building the foundational services developers use to create, deploy, and operate production-ready AI agents at scale. AgentCore Harness, launched at AWS Summit New York in June 2026, provides a managed agent loop that enables customers to define an agent through configuration rather than build and operate orchestration infrastructure themselves.
Customers specify their model, instructions, tools, skills, memory, and execution limits. AgentCore Harness manages the agent loop, secure execution environment, session isolation, identity, networking, observability, and deployment lifecycle. This allows developers to move quickly from an idea to a production agent while retaining the flexibility to integrate custom tools, models, and environments.
As a Software Development Engineer, you will design, implement, test, deploy, and operate core components across the Harness control plane and data plane. You will solve challenging distributed systems problems involving agent execution, session state, tool orchestration, security, scalability, reliability, and low-latency streaming. You will own features throughout their lifecycle, contribute to technical designs, and deliver high-quality solutions for customers.
Harness sits at the intersection of AgentCore Runtime, Memory, Gateway, Identity, Observability, Browser, Code Interpreter, and Evaluations. You will collaborate with engineers across these services to integrate capabilities, improve cross-service APIs, and reduce friction throughout the agent development lifecycle.
A central part of this role is making sophisticated agent capabilities dramatically easier to use. You will work backward from customer needs to turn complex infrastructure and orchestration tasks into intuitive configuration, sensible defaults, reusable templates, and guided workflows. Your work will help customers move from an idea to a secure, observable, production-ready agent in minutes while preserving the flexibility advanced builders need for customization.
You will help develop capabilities that make agent development faster and more accessible. This may include simplifying how customers connect tools and memory, building richer agent templates and skills, automating permissions and resource wiring, shortening the build-test-deploy loop, and creating safer ways to evaluate and release changes.
This is an opportunity to join a newly launched service at an early stage, contribute to its architecture and customer experience, and deliver capabilities with broad industry impact. Machine learning experience is helpful but not required. We are looking for strong engineers with distributed systems experience, sound technical judgment, a commitment to operational excellence, and the ability to deliver in a fast-moving environment.
Key job responsibilities
- Design, build, test, deploy, and operate highly available control plane APIs, data plane services, and managed agent loops for production AI agents.
- Build secure, isolated execution capabilities spanning tool orchestration, memory, context management, streaming, identity, networking, observability, and failure handling.
- Develop integrations between Amazon Bedrock and third-party models, AgentCore services, MCP servers, AWS Step Functions, Strands Agents, and customer-defined environments.
- Implement versioning, endpoints, progressive deployment, rollback, evaluation, and optimization capabilities that help customers safely improve agent behavior.
- Analyze and improve latency, scalability, availability, resource efficiency, and cost controls across multi-tenant workloads.
- Simplify the developer journey across APIs, SDKs, CLI, console, infrastructure as code, and local workflows by building cohesive abstractions, templates, defaults, and automation.
- Work directly with customers and partner teams to identify and reduce friction, accelerate time to production, and balance approachable experiences with the flexibility advanced builders need.
- Contribute to architectural decisions within Harness and collaborate with teams across AgentCore. Help raise the engineering bar through design reviews, operational readiness, code reviews, knowledge sharing, and mentoring.