E
Generative AI Engineer
EXL
4 hours ago
Full-time
On-site
Pittsburgh, Pennsylvania, United States
Role : Sr. AI Engineer with Semantic Kernel + Microsoft Agent Framework exp. ( Azure AI)
Location : Pittsburgh, PA
Work Type : Hybrid work setup (2-3 days a week work from office)
Hire Type : Full-Time with EXL
Experience : 7+ Years
Job Requirement: Migration Context - Must Have Knowledge Search is a containerized, enterprise-grade RAG platform on Azure AI Services with a Django/Python, deployed on OpenShift. Semantic Kernel is the current orchestration layer, and the SK → MS Agent Framework migration is a Q3 '26 roadmap item that unblocks Agentic Workflows and the Control Panel. The refactor re-platforms SK Native Plugins, Function Calling, Prompt Templates, Memory/Chat History, and Planners/Agents onto the Agent Framework's agents/workflows model.
Must-Have Skills: Semantic Kernel + Microsoft Agent Framework
— hands-on experience refactoring SK plugins, planners, function calling, and memory to Agent Framework agents/workflows (this is the critical, hardest-to-source skill) Python / Django
— to rebuild the orchestration layer without breaking existing API contracts, RBAC, and permissions Azure AI Services / RAG
— Azure AI Search (hybrid retrieval, semantic ranking), Azure OpenAI (inference + embeddings), preserving retrieval, reranking, prompt assembly, and citation paths
High-Priority Skills Model Context Protocol (MCP ) — for tool/connector integration and the Intelligence Layer channel OpenShift / containerization
— to deploy migrated microservices on OCP Observability & Evaluation
— Arize Phoenix + OpenTelemetry instrumentation, agent/agentic-system registration, and RAG quality/eval validation to confirm parity GitHub / CI-CD / DevSecOps
— PR-based workflows and vulnerability remediation
Medium-Priority Skills Redis Enterprise (caching/session state) and Azure APIM (gateway, rate limiting) The single most differentiated requirement is combined hands-on Semantic Kernel and Microsoft Agent Framework experience paired with Python/Django, since that is the exact intersection where the refactor happens. The remaining skills (Azure AI Search, RAG, OCP, observability) align with what the team already runs today.
Job Requirement: Migration Context - Must Have Knowledge Search is a containerized, enterprise-grade RAG platform on Azure AI Services with a Django/Python, deployed on OpenShift. Semantic Kernel is the current orchestration layer, and the SK → MS Agent Framework migration is a Q3 '26 roadmap item that unblocks Agentic Workflows and the Control Panel. The refactor re-platforms SK Native Plugins, Function Calling, Prompt Templates, Memory/Chat History, and Planners/Agents onto the Agent Framework's agents/workflows model.
Must-Have Skills: Semantic Kernel + Microsoft Agent Framework
— hands-on experience refactoring SK plugins, planners, function calling, and memory to Agent Framework agents/workflows (this is the critical, hardest-to-source skill) Python / Django
— to rebuild the orchestration layer without breaking existing API contracts, RBAC, and permissions Azure AI Services / RAG
— Azure AI Search (hybrid retrieval, semantic ranking), Azure OpenAI (inference + embeddings), preserving retrieval, reranking, prompt assembly, and citation paths
High-Priority Skills Model Context Protocol (MCP ) — for tool/connector integration and the Intelligence Layer channel OpenShift / containerization
— to deploy migrated microservices on OCP Observability & Evaluation
— Arize Phoenix + OpenTelemetry instrumentation, agent/agentic-system registration, and RAG quality/eval validation to confirm parity GitHub / CI-CD / DevSecOps
— PR-based workflows and vulnerability remediation
Medium-Priority Skills Redis Enterprise (caching/session state) and Azure APIM (gateway, rate limiting) The single most differentiated requirement is combined hands-on Semantic Kernel and Microsoft Agent Framework experience paired with Python/Django, since that is the exact intersection where the refactor happens. The remaining skills (Azure AI Search, RAG, OCP, observability) align with what the team already runs today.