AI Prompt Chaining & Task Routing

In this hands-on project, you will explore the fundamentals of AI orchestration and prompt chaining by building an intelligent system for automated document processing. The goal is to move beyond simple prompts and learn how to design multi-step AI workflows that can analyze, interpret, and structure complex information.

Using LangGraph, you will build a stateful AI workflow capable of task routing, conditional logic, and iterative refinement. Instead of relying on linear prompt sequences, you will implement flexible architectures where the AI can dynamically decide how to process different sections of a document—creating workflows that resemble how humans read, analyze, and synthesize information.

Along the way, you will use Docling to perform high-fidelity document parsing, ensuring that layouts, formatting, and structural elements are preserved during extraction. To organize and validate the extracted information, you will leverage Pydantic to transform raw text into structured, reliable data models suitable for downstream automation and system-to-system communication.

By completing this project, you will gain practical experience with modern AI workflow frameworks, learning how to build robust prompt chains, intelligent routing systems, and scalable document-processing pipelines. The skills you develop will help you design production-ready AI systems capable of turning unstructured data into structured insights.

• OpenAI Account - signup here: https://auth.openai.com/create-account. Free tokens are available!
• IDE of your choice - developed on VS code
• Modern Python Version >=3.10  - virtual environments recommended
• Developed on Windows - no known operating system specific dependencies however
• Requirements.txt provided - notebooks, docling, openai and LangGraph