Email AI Agent using LangGraph (LangChain)

Developed an autonomous ReAct-style AI agent using LangGraph, enabling LLM-driven decision-making workflows with stateful graph transitions for enhanced context and control.

• Integrated Google Gemini (via LangChain) as the LLM backbone, leveraging built-in tool calling support.
• Built two custom tools:
• find email: Retrieve a user’s email from a remote SQL database using only their first name.
• send email: Compose and send dynamic emails by inferring appropriate subject and body from the given prompt.
• Designed the agent’s control flow using LangGraph’s StateGraph, managing tool selection, execution, and context-aware message handling.
• Enhanced workflow reliability by combining LangChain’s structured outputs with LangGraph’s deterministic execution paths.