Lead / Manager Agentic AI Engineer - Claude Code & Codex

Key Responsibilities

• Agentic AI Development & Orchestration - Design and develop LLM-powered autonomous and semi-autonomous agents capable of executing complex, multi-step workflows. Build agent workflows using frameworks such as LangGraph, LangChain, Semantic Kernel, AutoGen, or similar technologies. Implement planning, task decomposition, tool selection, execution, observation, retry, and validation loops. Develop agents that can interact with enterprise applications, APIs, databases, and developer tools through natural language.
• Context and Harness Engineering: Design and implement context engineering strategies that provide agents with the right instructions, task context, application state, tools, and relevant information at the right time. Develop AI agent harnesses that manage agent state, tool access, permissions, execution workflows, guardrails, retries, and verification. Engineer repository and application context for AI coding agents such as Claude Code, OpenAI Codex, or similar platforms. Develop effective agent instructions, project context, coding guidelines, workflows, and automated verification mechanisms to improve agent reliability and developer productivity. Optimize context usage to reduce unnecessary token consumption, latency, and LLM costs.
• MCP & Tool Integration: Design and develop Model Context Protocol (MCP) servers and tools that enable agents to interact with enterprise applications and services. Integrate agents with Git, GitHub/GitLab, Artifactory, Slack, databases, APIs, CI/CD platforms, and other enterprise tools. Build secure tool-calling mechanisms with appropriate authentication, authorization, permissions, and human approval workflows. Develop reusable tools that enable agents to perform actions rather than simply generate responses.
• LLM & Generative AI Engineering: Integrate and orchestrate LLMs for reasoning, planning, content generation, code generation, and task execution. Work with commercial and open-source LLMs and select appropriate models based on quality, latency, cost, and task complexity. Implement prompt engineering and advanced context management strategies. Apply techniques such as structured outputs, function/tool calling, model routing, and model fallback strategies. Implement RAG where required, including document retrieval, embeddings, vector databases, reranking, and grounding.
• Agent Evaluation & Reliability: Design evaluation frameworks to measure agent task completion, tool-call accuracy, response quality, hallucination, reliability, and business outcomes. Implement LLM-as-a-Judge and automated evaluation pipelines for agentic and GenAI applications. Build regression testing and validation workflows for agent behavior. Implement guardrails, error handling, retry mechanisms, and human-in-the-loop controls for high-risk actions.