Agentic Workflow Design: Map end-to-end business processes to identify high-value opportunities for autonomous agent intervention in areas like Wealth Management, Risk Management, or Compliance.Autonomy & Guardrail Definition: Define task autonomy, decision thresholds, and escalation paths (Human-in-the-loop) to ensure agents operate within strict compliance and risk frameworks.Technical Specification: Translate business needs into technical designs for Agent Orchestration systems, including Third-Party LLM integration, RAG pipelines, and tool-calling capabilities via APIs.Goal-Oriented Requirements: Draft "Agent Personas" and behavioral specs, moving beyond simple user stories to define agent goals, constraints, and success metrics.Performance Evaluation: Lead the evaluation of agent reasoning and tool usage, utilizing frameworks like "LLM-as-a-Judge" to ensure decision quality and accuracy.
