Descripción de la oferta
We’re looking for a Tech Lead specializing in Agentic Gateway Architecture to lead the design and implementation of a secure, scalable integration layer for AI agents, tools, and multi-agent workflows. The focus is on MCP tool integration, semantic tool discovery, A2 A communication, identity federation, authorization, and gateway architecture within an enterprise AI platform. The project involves building an enterprise-grade AI Agent Platform from scratch, providing a standardized foundation for developing, deploying, orchestrating, securing, and observing AI agents across multiple teams and use cases. The platform will cover the full agent lifecycle, including orchestration, observability, security, governance, integrations, and evaluation. This is a greenfield initiative, giving engineers the opportunity to shape the architecture and make key technical decisions rather than maintain an existing solution. The client is a large global enterprise investing heavily in modern cloud, data, and AI capabilities to support secure, scalable, and highly automated solutions. Requirements: Model Context Protocol (MCP) — server aggregation, tool discovery, tool invocation lifecycle Agentic workflow patterns — agent-to-tool, agent-to-agent communication (A2 A), multi-agent orchestration Semantic tool discovery — embedding-based search for agents to find tools by capability/intent Agentic authorization — policy enforcement on tool calls and agent interactions (Cedar policy language) Agent identity federation — JWT-based cross-agent authentication (Entra Agent ID or equivalent) AWS Bedrock Agent Core Gateway — MCP target registration, HTTP target routing, inference target configuration Experience: 5+ years building AI/ML platforms or agentic systems Designed tool integration layers for AI agents (function calling, tool use) Implemented Model Context Protocol or similar agent-tool protocols Enterprise security reviews for AI systems (authorization, data access, audit) Nice to have: Built multi-agent systems with agent-to-agent communication Semantic search using embeddings (FAISS, Pinecone, vector databases) AWS Bedrock Agents or Agent Core hands-on experience Agentic gateway topology design (central vs federated, tool visibility scoping) Credential vending for agents (3 LO OAuth, per-tool authentication) Responsibilities: Lead the architecture and implementation of an agentic gateway connecting AI agents with enterprise tools and services. Design MCP-based integration patterns, including server aggregation, tool discovery, and tool invocation lifecycle. Define and implement agent-to-tool and agent-to-agent communication patterns, including multi-agent orchestration. Design semantic tool discovery using embeddings to enable agents to identify tools based on capability and intent. Establish authorization and policy enforcement for tool calls and agent interactions using Cedar policies. Design agent identity and authentication patterns using JWT-based federation and Entra Agent ID or equivalent solutions. Define gateway topology, including centralized vs federated models, tool visibility, and access scoping. Integrate and configure AWS Bedrock Agent Core Gateway, including MCP target registration, HTTP routing, and inference targets. Ensure the architecture meets enterprise requirements for security, authorization, data access, and auditability. Provide technical leadership and architectural guidance for agentic platform and gateway implementation.
