Description:

This hands-on course introduces modern AI-assisted software development using OpenAI Codex. Participants will learn how to move beyond simple prompt-based coding into structured, agentic workflows that use context engineering, spec-driven development, and multi-session orchestration.

By the end of the day, learners will be able to design and build real software systems using Codex as an active development partner—capable of writing, refactoring, planning, debugging, and extending code across local and cloud environments.

The course emphasizes practical engineering patterns such as context management, work tree-based parallel development, MCP integrations, and task decomposition for scalable AI-driven software delivery.

Duration:

1 Day

Course Code: BDT 620

Learning Objectives:

After this course, you will be able to:

1. Engineer: Use OpenAI Codex effectively to generate, refactor, and debug production-quality code
2. Design: Apply context engineering principles to improve reliability and control of AI-generated outputs
3. Build: Develop full applications using spec-driven development with Codex as a collaborative agent
4. Orchestrate: Manage local and cloud-based execution workflows for AI-assisted development tasks
5. Extend: Integrate MCPs and external tools to expand Codex capabilities
6. Scale: Use session forking and git work trees to parallelize development streams

Software engineers, AI engineers, data professionals, technical product builders, and developers transitioning into AI-assisted development workflows.

Basic programming experience (preferrable Python). Familiarity with Git/GitHub is recommended but not required.

Course Outline:

1. Foundations of Agentic Coding with Codex
   1. Introduction to Codex as an AI Software Engineering Agent
   2. Prompting vs Context Engineering: Why context wins
   3. Anatomy of a Codex-driven development loop (plan → build → refine → test)
   4. Lab: Build your first multi-step Codex-assisted feature
2. Context Engineering & Spec-Driven Development
   1. Designing effective specifications for AI coding agents
   2. Iterative development: feature expansion and refactoring loops
   3. Context window management strategies for large projects
   4. Lab: Convert a vague product idea into a structured Codex spec
3. Building Applications with Codex
   1. Rapid application scaffolding using Codex
   2. Grounding agents in proprietary documents, datasets, and knowledge bases
   3. Debugging with AI: error tracing and self-healing code patterns
   4. Code quality control: enforcing structure, style, and patterns
   5. Lab: Build a small end-to-end application using Codex workflows
4. Agentic Engineering Workflows
   1. From scripts to agents: evolving Codex usage patterns
   2. Task decomposition and autonomous execution strategies
   3. Forking sessions for parallel experimentation
   4. Lab: Parallel feature development using forks and work trees
5. Extending Codex with MCPs & External Tools
   1. Understanding MCPs (Model Context Protocol integrations)
   2. Connecting Codex to external APIs and toolchains
   3. Tool-augmented coding workflows
   4. Lab: Extend Codex with external service integration
6. Productionizing AI-Assisted Development
   1. Structuring AI-generated code for production readiness
   2. Version control strategies for AI-assisted codebases
   3. Deployment considerations for AI-built applications
   4. Lab: Build and package a Codex-driven application

Training material provided: Yes (Digital format)

Hands-on Lab: Students will need a OpenAI account and we will use VS Code