Projects
LLM_Integration_and_Workflows.ipynb
This project builds on Module 1 by showcasing a modular agent framework using OpenAI’s GPT-4o-mini for multi-step problem-solving. Created for MSDS 442 at Northwestern, it follows a structured reasoning loop, Thought, Action, PAUSE, Observation, and integrates tools like LangChain for dynamic data retrieval and calculations. The project highlights how well-structured workflows enable AI agents to reason clearly and handle complex tasks beyond simple Q&A.
Features
- LLM Integration: Uses OpenAI’s GPT-4o-mini model, integrated into a custom agent framework for conversational interactions and problem-solving.
- Agent Workflow: Implements a four-step reasoning loop (Thought, Action, PAUSE, Observation), where the agent reflects, performs actions, and then outputs results based on observation.
- Tool Use: Dynamically retrieves or calculates information based on user input.
- Transparent Output: Shows the agent’s logic and steps at each stage.
- Complex Task Handling: Solves multi-step problems through sequential reasoning.
Key Insight
LLMs become significantly more powerful when paired with structured agent workflows, enabling clear, multi-step problem solving, not just single-response answers.