Course Overview

Course Overview#

Course Description#

This course provides hands-on experience with generative AI and agentic technologies and their practical applications in quantitative finance. Students will start by mastering modern AI development tools like Cursor editor and GitHub Copilot, then progress through understanding AI tool use, building autonomous agents, and implementing sophisticated systems like RAG. The course emphasizes real-world implementation using cutting-edge AI frameworks, covering everything from LLM fundamentals and the Model Context Protocol to building custom AI applications and deploying them in production environments. By the end of the course, students will be proficient in creating production-ready AI applications that can transform financial workflows and decision-making processes through intelligent automation and enhanced data analysis.

Topics#

AI Development Tools & Copilots#

  • Modern AI Development Environments: Introduction to Cursor editor and GitHub Copilot in Visual Studio Code

  • AI-Assisted Programming: Understanding how AI copilots enhance coding workflows and productivity

  • Prompt Engineering: Crafting effective prompts for financial analysis and trading

  • Tools and MCP Servers for AI Copilots and Agents

AI Tool Use & Agentic Workflows#

  • Command Line AI Tools: Introduction to Claude Code and OpenAI Codex CLI for terminal-based AI assistance

  • Tool Use Fundamentals: How LLMs invoke external tools and execute commands

  • Web Search as a Tool: Integrating real-time information retrieval into AI applications

  • Model Context Protocol (MCP): Building custom MCP servers and understanding the protocol architecture

  • Agentic AI Systems: Moving from simple tool use to autonomous AI agents that can execute complex workflows using LangGraph and LangChain

Large Language Model Fundamentals#

  • Deep Dive into LLMs: Understanding tokenization, next-token prediction, and model architecture

  • Training for Tool Use: How LLMs learn to use tools and the mechanics of tool invocation

  • Tool Use vs. MCP: Comparing different approaches to AI tool integration

Challenges of LLM’s for Financial Data#

  • Transformers and self-attention: Challenges with respect to financial market data

  • Case Study: “Virtue of Complexity” Debate

Retrieval-Augmented Generation (RAG)#

  • RAG as Tool Use: Understanding RAG as a sophisticated tool for information retrieval

  • Vector Databases: Storing and querying financial document embeddings using tools like Pinecone, Weaviate, and Haystack

  • Document Processing: Chunking strategies and content filtering for financial data

  • RAG Application Architecture: Building complete knowledge retrieval systems with LangChain and Haystack

AI Application Development & Deployment#

  • Local vs. Cloud AI Development: Exploring local LLM deployment and OpenAI API integration

  • Custom AI Chat Applications: Building personalized chatbot interfaces (e.g., T3 Chat example)

  • Model Selection & Integration: Choosing appropriate AI models for financial tasks

  • Production Deployment: Deploying AI applications with proper monitoring and governance

AI Governance & Evaluation#

  • Response Quality Assessment: Evaluating AI outputs and implementing guardrails

  • Performance Benchmarking: Evaluating AI model performance using quantitative metrics

  • Response Evaluation: Implementing frameworks for testing and validating AI applications

  • AI Safety & Compliance: Building safeguards for financial AI systems

  • Production Monitoring: Tracking performance and maintaining AI applications in production

Prerequisites#

  • FINM 32900: Full Stack Quantitative Finance (or consent of instructor)

  • Basic Python programming skills

  • Familiarity with machine learning concepts

  • Understanding of financial markets and instruments

Learning Outcomes#

By the end of this course, students will be able to:

  • Use modern AI development tools like Cursor and GitHub Copilot effectively

  • Build, fine tune, and deploy AI applications that can use tools and execute commands

  • Understand the fundamentals of how LLMs work and how they use tools. Students will learn how to build their own Model Context Protocol (MCP) servers

  • Implement RAG systems for financial knowledge retrieval

  • Create custom AI applications using local and cloud-based LLMs

  • Deploy production-ready AI applications with proper governance and monitoring

  • Apply industry best practices for AI development in regulated financial environments

Contents