Lead AI Engineer (Agentic Systems)

Grade Level (for internal use): 11 Lead AI Engineer (Agentic Systems)   Role Summary   As the Lead AI Engineer (Agentic Systems), you will   help   architect and build the organization’s next generation of autonomous AI workflows. This is a multidisciplinary technical role   operating   at the intersection of Software Engineering, Data Engineering, and Machine Learning   Engineering . You will move beyond simple "chatbots" to design production-grade Agentic Systems: intelligent applications capable of reasoning, planning, and executing complex tasks autonomously.

Agentic Systems Architecture & Core Engineering   Architect & Build Multi-Agent Workflows: Lead the hands-on design and coding of stateful, production-grade agentic systems using Python and orchestration frameworks like   LangGraph ,   CrewAI , or   AutoGen .   Agent-to-Agent (A2A) Communication: Design and implement robust A2A protocols enabling autonomous agents to collaborate, hand off sub-tasks, and negotiate execution paths dynamically within multi-agent environments.   State Management & Orchestration: Engineer robust control flows for non-deterministic agents; implement complex message passing, memory persistence, and interruptible state handling to support long-running autonomous tasks.   Tool Interface Design (MCP): Implement and standardize the Model Context Protocol (MCP) to create universal interfaces between agents, data sources, and operational tools, ensuring modularity and scalability.   Model Integration & Optimization:   Utilize   proxy services ( i.e.   LiteLLM )   to manage model routing and fallback strategies;   optimize   context windows and inference costs across proprietary and open-source models.     Production Deployment: Containerize agentic workloads using Docker and orchestrate deployments on Kubernetes; leverage AWS   AgentCore   or similar cloud-native services for scalable infrastructure.   Data Engineering & Operational Real-Time Integration   Build Agent Data Pipelines: Write and   maintain   high-throughput ingestion pipelines (using Databricks or Python-based ETL) that transform raw operational signals into structured context for agents.   Real-Time Context Injection: Ensure agents have access to "operational real-time" data (seconds/minutes latency) by   optimizing   retrieval architectures and vector store performance.     Cross-Functional Engineering: Act as the technical bridge between Data Engineering and AI teams; translate complex agent requirements into concrete data schemas and pipeline specifications, while stepping in to resolve hands-on bottlenecks in data availability.   Observability, Governance & Human-in-the-Loop   LLMOps   & Tracing: Implement comprehensive observability using tools like   Langfuse   to trace agent reasoning steps,   monitor   token usage, and debug latency issues in production.     Safety & Control Frameworks: Design hybrid execution modes ranging from Human-in-the-Loop (HITL) for sensitive operations to fully autonomous execution; build "break-glass" mechanisms and guardrails for automated decision-making.     Evaluation & Reliability:   Establish   technical standards for testing non-deterministic outputs; automate evaluation pipelines to measure agent accuracy, hallucination rates, and drift before deployment.     Technical Leadership & Strategy   Technical Roadmap Definition: Partner with Product and Engineering leadership to scope feasibility for autonomous projects; define the "Agentic Architecture" roadmap.   Mentorship & Standards: Define code quality standards, architectural patterns, and PR review processes for the AI engineering team; upskill team members on the latest agentic frameworks and methodologies.   Innovation: Proactively prototype with emerging tools (e.g., new reasoning models, graph-based RAG) to solve high-value business problems, moving successful experiments into the production