Prompt Engineering, LLM APIs, GitHub Copilot & AI Code Review
AI isn't replacing senior developers — but developers who use AI effectively are replacing those who don't. This intensive course gives experienced developers the practical skills to integrate LLMs into their workflows, automate code review, and build AI-powered features using the OpenAI and Anthropic APIs.
How LLMs work at an engineer's level — tokens, context windows, temperature, sampling. Comparing models: GPT-4o, Claude Sonnet/Opus, Gemini. Choosing the right model for the job and understanding cost vs capability tradeoffs.
System prompts, few-shot examples, chain-of-thought, structured output (JSON mode), role prompting and prompt chaining. Learn what makes prompts reliable versus brittle — and how to test them.
Tree-of-thought, self-consistency, retrieval-augmented prompting, constitutional AI patterns, and prompt compression. Practical strategies for production prompts where consistency matters.
Chat completions, streaming responses, function calling (tool use), embeddings and semantic search. Building a simple Python service that wraps GPT — with error handling, retries, rate limit management and cost tracking.
The Messages API, multi-turn conversations, tool use, streaming and prompt caching. Comparing Claude and GPT API patterns, migrating between providers, and building provider-agnostic wrappers.
IDE integration (VS Code, JetBrains), autocomplete strategies, Copilot Chat for inline refactoring and explanation, slash commands, and Copilot for pull request summaries. Making Copilot work for you rather than against you.
Using LLMs to generate, review and explain code. Prompting patterns for security review, test generation, documentation and legacy code refactoring. Building an AI code-review step into your CI pipeline.
CLI-based agentic development: CLAUDE.md context files, custom slash commands, skills, and MCP (Model Context Protocol) server integration. Using Claude Code to scaffold features, run tests and manage multi-file changes autonomously.
Embedding-based retrieval, vector stores, building a simple RAG pipeline. Function-calling agents, multi-step reasoning, tool orchestration — and when to use agents versus simpler prompting patterns.
Testing LLM outputs: evals, benchmarks, guardrails and hallucination mitigation. Capstone: build and demo an AI-powered developer tool (code reviewer, documentation generator or internal chatbot) using the APIs covered in the course.
No time wasted on basics. This course is designed for experienced developers — we move fast, go deep, and focus on patterns you can apply in a real production codebase from day one.
Every module includes working code. You'll call real APIs, build real integrations and leave with a portfolio of reusable AI utilities and prompt templates.
Available as a 5-day open course in Johannesburg, a 100% online cohort, or a 20-hour condensed workshop delivered at your offices anywhere in South Africa.
Receive a Code College certificate on completion. Training qualifies as skills development under the B-BBEE scorecard and counts toward your SDL levy spend.
Not necessarily. Familiarity with REST APIs and basic scripting in any language is helpful. The course is designed for experienced developers — if you have 2+ years of professional development experience in Java, C#, PHP, JavaScript or another language, you will follow the material comfortably.
This course is built specifically for software developers, not data scientists or business analysts. The focus is on integrating AI into a developer workflow — calling LLM APIs, using GitHub Copilot in a real codebase, automating code review, and building AI-powered tools — not on statistics or model training.
Yes. A 20-hour condensed workshop is available for corporate teams, covering the core modules: Prompt Engineering, OpenAI and Claude APIs, GitHub Copilot, and AI-Assisted Code Review. It runs as a 2.5-day intensive, delivered onsite at your offices anywhere in South Africa. This is priced per group rather than per delegate — contact us for a quote based on your team size and location.
Yes. Code College is a B-BBEE Level 4 contributor and training spend qualifies as skills development under the B-BBEE scorecard. The course fee also counts toward your SDL (Skills Development Levy) spend for SETA reporting purposes.
You will work hands-on with the OpenAI API (GPT-4o), the Anthropic Claude API, GitHub Copilot, and Claude Code for agentic workflows. You will also explore RAG (Retrieval-Augmented Generation) patterns and build a capstone AI-powered developer tool — a code reviewer, documentation generator, or internal chatbot.
Yes. Corporate onsite delivery is available anywhere in South Africa. Contact us to discuss scheduling, group rates, and customising the content to your team's tech stack.
Book the AI Tools for Developers course — 40-hour open cohort or 20-hour corporate workshop. Johannesburg and online.