Senior AWS Agentcore Platform Engineer

Senior AWS Agentcore Platform Engineer

Position Type: Contract to hire after initial 6 months

Location: Reading, PA or Exton, PA (Hybrid 2-3 days a week from office)

1. Observability & Distributed Tracing

• Gap Analysis: Assess AWS CloudWatch, X-Ray, Bedrock logging, and AgentCore traces against agentic workflow requirements; produce a comprehensive gap analysis and lead the setup of observability within Dynatrace.
• Validation Pipelines: Design and implement post-deployment validation pipelines for agents and Model Context Protocol (MCP) servers, ensuring deployment health and successful tool registration.
• Tracing & Logging: Implement distributed tracing and structured logging to capture LLM decision logic, tool selections, sub-agent calls, and MCP interactions.
• Architecture Strategy: Evaluate LangFuse and LiteLLM proxies against AWS-native solutions; deliver a target-state observability architecture recommendation.

2. Cost Tracking & TCO (Total Cost of Ownership)

• Taxonomy Expansion: Extend tagging taxonomy to capture costs across agent runtimes, MCP servers, vector databases, and Bedrock token consumption per namespace.
• Cost Modeling: Design a granular cost visibility model to aggregate expenses for agents, MCPs, and LLM tokens by team and department.
• Dashboards & Alerting: Build CloudWatch (or equivalent) dashboards for per-team spending; configure AWS Budgets with proactive alerting thresholds.
• Automation: Automate cost reporting via email and Microsoft Teams, incorporating anomaly detection rules to identify spend spikes.

3. Monitoring & Incident Management

• Alerting Framework: Define and implement P1–P4 alerting rules covering deployment failures, runtime errors, tool invocation failures, and MCP connectivity issues.
• Incident Integration: Integrate alert notifications with Microsoft Teams and email, utilizing resource ownership tags for intelligent routing.
• Operational Excellence: Author detailed runbooks for every alert; publish and maintain these in Confluence to facilitate developer self-service resolution.
• Stack Evaluation: Compare AWS-native vs. third-party monitoring stacks to deliver a long-term recommendation aligned with the broader observability architecture.

4. Security & Governance

• Risk Assessment: Evaluate current IAM and tagging strategies for multi-team isolation; identify scalability gaps and potential security risks.
• Policy Engines: Assess the Cedar policy engine (AgentCore) for fine-grained tool access control and document gaps for enterprise-scale deployment.
• Identity Architecture: Design a scalable Attribute-Based Access Control (ABAC) identity model to ensure multi-team isolation without IAM policy sprawl; deliver production-ready Terraform modules.