Building Agentic AI Systems with AWS Bedrock AgentCore for Unified Talent Marketplaces

DHI Group faced the challenge of quickly moving generative AI proofs of concept to production. The traditional software development lifecycle (SDLC) was too slow for their accelerated pace. To overcome this, they partnered with AWS using a structured Hackathon Acceleration Package (HAP) to validate technical feasibility, build organizational AI literacy, and produce shippable code rapidly.

Agentic AI Architecture for Unified Systems

The winning hackathon solution, "ClearanceJobs MCP Server + AgileATS", showcased a modern agentic AI architecture pattern designed to unify disparate systems through intelligent automation. This architecture uses the Model Context Protocol (MCP) to expose system capabilities as 'tools' that an AI agent can orchestrate, providing a single, intelligent interface for complex multi-system workflows.

Key Architectural Components

• Amazon Bedrock AgentCore (Orchestration Layer): Provides the full agent infrastructure, including Agent Runtime for session management and reasoning loops, and a Gateway for tool discovery and routing. It handles secure authentication to downstream MCP servers.
• MCP Server Lambda (Tool Layer): A Lambda function exposing system capabilities as discrete tools (e.g., `search_candidates`, `get_candidate`). It connects to backend systems via a NAT gateway with WAF-allowlisted egress for security.
• ProfileLookup Lambda (External Enrichment): A separate Lambda function for enriching candidate data with external information, such as GitHub profiles, by calling external APIs.
• Foundation Model (Reasoning Layer): Anthropic's Claude 3.5 Haiku in Amazon Bedrock interprets natural language commands, decomposes requests into tool calls, and synthesizes responses.
• AgentCore Memory (Context Layer): Persists session state and recruiter preferences across conversations, allowing the agent to recall context and patterns.

Security and Networking Considerations

The solution spans two AWS accounts for enhanced security and separation of concerns: an AgileATS account for AgentCore components and the foundation model, and a ClearanceJobs account for the MCP Server and ProfileLookup Lambda functions within a VPC. Communication between AgentCore and the ClearanceJobs account occurs over HTTPS using MCP with bearer authentication and custom headers for tenant identification, ensuring secure and authorized data exchange.