Agentic AI Manager

Key Responsibilities

1. Agent Design & Architecture

• Design end-to-end AI agent architectures including single-agent and multi-agent systems using patterns such as ReAct, Reflection Pattern.
• Define agent goals, tools, MCP, memory strategies (in-context, external, semantic, episodic), and failure-handling logic.
• Select the appropriate agent pattern and framework based on use-case complexity and enterprise constraints
• Build agentic RAG systems combining vector search, semantic retrieval, and LLM reasoning for knowledge-grounded responses

2. Development & Integration

• Develop AI agents using Python-based frameworks for code-first builds
• Build and configure agents on Microsoft Copilot Studio with custom connectors, adaptive cards, and Power Automate flow integrations
• Integrate agents with enterprise APIs: SAP, Snowflake (Cortex), Microsoft Graph API, SharePoint, Teams, and other M365 services etc.
• Implement MCP (Model Context Protocol) servers to expose tools and data sources to agents in a standardized, composable manner
• Build tool-calling layers that allow agents to interact with databases, ERPs, and third-party APIs in real time
• Design human-in-the-loop checkpoints for approval workflows, exceptions, and escalation scenarios

3. RAG & Knowledge Management

• Design and implement Retrieval-Augmented Generation (RAG) pipelines: document ingestion, chunking, embedding, vector store indexing, and context-aware retrieval
• Work with vector databases (Pinecone, Azure AI Search, ChromaDB) to build semantic search capabilities for enterprise knowledge bases
• Implement agentic RAG patterns including query routing, self-reflection, and query reformulation for higher-quality responses

4. Evaluation, Testing & Observability

• Build evaluation frameworks to measure agent accuracy, task completion rate, hallucination rate, and latency
• Implement observability and tracing to debug and improve agent behavior in production
• Define and monitor KPIs for deployed agents; establish feedback loops for continuous improvement
• Conduct structured prompt engineering and system prompt optimization to improve reliability and output quality

5. Governance, Safety & Responsible AI

• Implement guardrails, content filters, and safety mechanisms to ensure agents behave within defined boundaries
• Apply responsible AI principles: fairness, transparency, explainability, and data privacy compliance
• Document agent logic, decision flows, tool schemas, and known limitations for audit and knowledge transfer

6. Collaboration & Stakeholder Management

• Partner with business process owners to translate operational problems into agent design briefs
• Collaborate with IT, data engineering, and enterprise architecture teams for system integration and cloud deployment
• Present agent capabilities, prototypes, and performance reports to senior leadership including CIO-level audiences
• Mentor junior team members and Management Trainees on agent concepts, tools, and development best practices