Agentic AI Architect
Soumik SahaUnited StatesSENIORPosted 48m ago
AI/ML platform systemsLLM orchestration frameworksLangChainLangGraphSemantic Kerneldistributed systems designmicroservices architectureevent-driven patternsAPI designRESTgRPCPythonJavaGoTypeScriptagentic AI systemsLLM-powered applicationsprompt engineeringtool-use patternsRAG pipelinesKubernetescontainer orchestrationcloud platformsAWSAzureGCPMLOpsLLMOpsvector databasesPineconeWeaviatepgvectorknowledge graphsTOGAF
About the Role
We are looking for an Agentic AI Architect to join the Agentic System Layer (ASL) team at American Airlines. This role involves defining and driving the technical architecture for the company's agentic AI platform.
Key Responsibilities
• Designing and evolving the architecture for multi-agent orchestration systems, tool-use frameworks, and LLM integration pipelines
• Establishing patterns for agent reliability, observability, and guardrails at production scale
• Leading technical design reviews and producing architecture decision records (ADRs)
• Collaborating with ML engineers and software engineers to ensure platform components are scalable, secure, and maintainable
• Evaluating and integrating emerging agentic AI frameworks (e.g., LangGraph, CrewAI, Semantic Kernel, AutoGen)
• Defining API contracts, data flow patterns, and integration standards across the AI platform ecosystem
• Mentoring engineers on best practices for building production-grade AI systems
Requirements
• 10+ years software architecture experience with at least 3 years designing AI/ML platform systems, including hands-on experience with LLM orchestration frameworks (LangChain, LangGraph, Semantic Kernel, or similar)
• Deep expertise in distributed systems design, microservices architecture, event-driven patterns, and API design (REST/gRPC), with strong proficiency in Python and at least one of Java/Go/TypeScript
• Production experience building and deploying agentic AI systems or LLM-powered applications at scale, including prompt engineering, tool-use patterns, RAG pipelines, and agent reliability/observability