Key skills
AILLMClaudeAnthropicAgenticAutoGenLangGraphCommunication
About this role
Agilyti is seeking a senior Lead Agentic AI Architect to design and guide the development of a production-grade, agent-based orchestration and simulation platform supporting complex clinical workflows. The role involves architecting multi-agent systems and optimizing the use of Anthropic Claude for high-stakes workflows.
Responsibilities:
- Design and implement multi-agent workflows to support complex operational and analytical use cases
- Architect orchestration layers that coordinate LLM reasoning, deterministic logic, and tool invocation
- Define long-context and memory strategies (100K+ token workflows)
- Build systems that balance rapid experimentation with long-term architectural integrity
- Lead the architectural use of Claude 3.x across reasoning, tool use, summarization, and structured output
- Optimize prompts and context engineering specifically for Claude’s reasoning strengths
- Implement Claude tool-use patterns and safety boundaries
- Define guardrails and reliability frameworks for high-stakes workflows
- Design agents capable of:
- Tool invocation
- External API integration
- Memory/state management
- Retry and fallback logic
- Agent-to-agent communication
- Implement orchestration patterns using frameworks such as LangGraph, CrewAI, AutoGen, Semantic Kernel, LiteLLM, or custom-built infrastructure
- Design and deploy Model Context Protocol (MCP) or equivalent tool-integration systems
- Architect deterministic + reasoning hybrid workflows
- Lead fine-tuning efforts (instruction tuning, LoRA, domain adaptation)
- Build evaluation pipelines to measure model performance
- Implement human-in-the-loop and validation feedback loops
- Optimize for latency, token efficiency, and reliability
- Define benchmarking frameworks for model evaluation and iteration
- Translate ambiguous business requirements into scalable AI architectures
- Present system design decisions to executive and technical stakeholders
- Operate with high ownership in early-stage, high-ambiguity environments
- Mentor engineering teams and guide implementation decisions