01Overview
ROLE SUMMARY
Leads the design of multi-agent systems across products. Owns MCP and A2A integration standards, agent orchestration and state management patterns, and production guardrails. Primary technical authority for agentic system architecture for the Product.
KEY RESPONSIBILITIES
- Architect multi-agent workflows, semantic routers, and autonomous systems using LangGraph, AutoGen, or CrewAI
- Design and govern MCP and A2A frameworks for secure agent-to-tool and agent-to-agent communication
- Implement orchestration layers managing agent memory, context, planning, and task execution
- Define guardrails, safety policies, and hallucination-prevention protocols for production agentic AI
- Design observability pipelines to trace, evaluate, and monitor multi-agent decision chains at scale
- Mentor Senior Agentic AI Engineers and MCP Integration Engineers
TECHNICAL SKILLS
- Languages & Frameworks: Python, TypeScript, Java; LangGraph, AutoGen, CrewAI, LlamaIndex
- AI Coding Tools: Claude Code, GitHub Copilot, OpenAI Codex, Cursor
- Protocols: MCP, A2A connector architecture and agent card design
- LLMs & GenAI: RAG, hybrid retrieval, prompt engineering, vector databases (Pinecone, pgvector, Weaviate, Qdrant)
- Structured Outputs & Function Calling: JSON schema enforcement, tool schema authoring
- Prompt Caching: Anthropic / OpenAI caching for latency and cost optimisation
- Sandbox Execution: E2B, Modal safe agent code execution workplace design
- AI Security: Guardrails AI, NeMo Guardrails, prompt injection prevention, access controls
- Observability: LangSmith, Arize Phoenix; agent harness testing, evals (Ragas, DeepEval, HELM)
- Cloud AI: AWS Bedrock, Azure AI Foundry, GCP Vertex AI
- HITL Design: human escalation paths and graceful agent handoff patterns
- Agent Orchestration: Task Decomposition, Subagent Delegation, Progressive Context Loading
- Interactive Planning & Grill-Me Prompting: constraint validation before any code or config is committed
NICE TO HAVE
- AWS / Azure / GCP Solutions Architect or AI certifications
- MLOps tooling: MLflow, Kubeflow, DVC
- Open-source contributions to LangGraph, AutoGen, or CrewAI
- EU AI Act compliance and responsible AI audit experience .
