AgentCore Platform Engineer

• We’re hiring AgentCore Platform Engineers to build the agent-facing layer of the platform. That covers how agents run, remember and work together, and how they safely reach the internal systems the business already runs on
• Each engineer will have a main focus in one of the two areas below and work closely across both. What you build becomes the template other teams copy, so quality and clarity matter as much as speed. This is a named-team engagement, so the person we propose is the person who starts
• Agent Runtime & Orchestration: Build runtime patterns, memory strategies, orchestration and multi-agent handoffs, and the reference implementations other teams copy
• Gateway, Tools & Integrations: Turn internal APIs, Lambdas and legacy systems into governed tools through AgentCore Gateway, including MCP server development
• Runtime Patterns: Define how agents are structured, packaged and run on AgentCore Runtime, including sessions, state, error handling and timeouts
• Memory Strategies: Design short-term and long-term memory patterns using AgentCore Memory, and guide teams on when to use which
• Orchestration and Multi-Agent Handoffs: Build patterns for planning, tool use, supervisor/worker setups and handoffs between agents, with clear boundaries and fallbacks
• Tool Onboarding: Expose internal REST APIs, Lambda functions and legacy services as agent tools through AgentCore Gateway, using OpenAPI specs and Lambda targets
• MCP Server Development: Design and build Model Context Protocol (MCP) servers for systems that need custom integration, with clean tool definitions and schemas
• Governance and Access: Set up authentication and authorization for agents and tools (AgentCore Identity, OAuth, IAM) so agents only reach what they’re allowed to
• Legacy Integration: Work with system owners to wrap older or on-premises systems safely, handling rate limits, data mapping and error cases
• Reference Implementations and Standards: Build production-quality reference agents, starter kits and a repeatable process for adding, versioning and reviewing tools
• Evaluation and Guardrails: Build in evaluation, Bedrock Guardrails, contract tests and safety checks so agents and tools behave predictably before going live
• Enablement: Run walkthroughs, write docs and support the teams onboarding to the platform

We know production experience with Bedrock AgentCore is rare. We care more about strong AWS fundamentals and real hands-on work with agent frameworks than about whether you’ve used AgentCore by name. If that’s you, practical experience with LLM tool/function calling and handling model outputStrong AWS fundamentals: IAM, Lambda, API Gateway, containers, DynamoDB/S3, VPC connectivity and Secrets ManagerClean, well-tested code that other engineers will read and reuseClear communication and comfortable teaching patterns to other teamsDeep hands-on strength in at least one of the two focus areas, and working knowledge of the otherUnderstanding of agent memory, retrieval (RAG) and context management trade-offsHands-on experience building LLM agents with at least one agent framework (e.g., Strands Agents, LangGraph, LangChain, CrewAI, OpenAI Agents SDK, LlamaIndex)5+ years of software engineering experience, with strong Python (TypeScript is a plus)Experience designing and consuming REST APIs and OpenAPI specs, and auth patterns such as OAuth 2.0 and OIDCYou pick up new AWS services quickly and learn by buildingExperience in pharma, life sciences or another regulated industry with data access controlsYou work well inside a client team and enjoy helping other engineers succeedHands-on experience with Amazon Bedrock AgentCore (Runtime, Memory, Gateway, Identity). A strong plus, but not requiredYou think about security and access control before anything goes liveExperience building or running MCP servers in a real environmentPrior experience and connections in the Life Sciences industry is preferredExperience with multi-agent systems in production, including observability and failure handlingYou’re honest about what agents can and can’t do todayYou’ve built agents that actually work, and you know where they tend to breakYou have prior experience in consultingYou care about patterns other people can copy, not just clever one-offsExperience with LLM evaluation tools and approachesExperience integrating with enterprise or legacy systems, or with API management platforms (e.g., MuleSoft, Apigee, Kong)