smart_toyIntermediate Course 3 of 4 workspace_premiumCertificate Included

AI Development

Stop using other people's AI products. Build your own. Connect Claude, OpenAI, and Gemini APIs. Run open-source models locally with Ollama. Implement RAG with vector search. Every module ships working code on your own server.

starstarstarstarstar
5.0 · Instructor-led · In-class practical schoolEnrol — R5 499
schedule24 hours of content layers12 modules · 56 lessons codePHP & JavaScript translateEnglish updateUpdated May 2026 all_inclusiveLifetime access lockSecurity-first approach

What you'll learn

By the end of this course you'll be able to:

check_circleConnect to Anthropic, OpenAI, Gemini, and xAI APIs from PHP and JS
check_circleBuild SSE-streaming AI chat interfaces from scratch
check_circleRun Llama, Mistral, and Phi locally with Ollama
check_circleImplement RAG (Retrieval Augmented Generation) with vector search
check_circleParse and analyse uploaded documents with AI (PDF, DOCX, CSV)
check_circleBuild a prompt template engine with variable injection
check_circleGenerate images with DALL-E 3 and Stable Diffusion
check_circleSecure AI apps: encrypt API keys, prevent prompt injection
check_circleAdd AI features to existing PHP applications
check_circleAudit and log every AI call for compliance and debugging

You'll work with these providers throughout:

🟤Anthropic Claude 🟢OpenAI GPT-4o 🔵Google Gemini 🟣xAI Grok 🦙Ollama / Local 🤗HuggingFace

Course Curriculum

12 modules · 56 lessons · 24 hours

Every module ends with a working app feature deployed on your server.

01 AI APIs — How They Work 4 lessons · 70 min expand_more
  • play_circleREST APIs, JSON, and HTTP in plain English Free Preview
  • play_circleAPI keys, authentication, and rate limits Free Preview
  • lockThe messages array: roles, turns, and context windows 20 min
  • lockModule project: Your first API call from PHP 18 min
code
Lab Outcome
A PHP script on your server making a live API call to Claude and returning a response — no library, just cURL and JSON.
02 Building a PHP AI Chat Interface 5 lessons · 95 min expand_more
  • lockSSE (Server-Sent Events): streaming responses live 20 min
  • lockBuilding the PHP streaming endpoint 22 min
  • lockJavaScript EventSource and rendering streamed tokens 18 min
  • lockPersisting conversation history in a session and database 16 min
  • lockModule project: Fully functional streaming chat UI 19 min
code
Lab Outcome
A Claude-powered chat interface — tokens stream live, history persists in MariaDB, and it renders markdown in the browser.
03 Multi-Provider Dispatcher 4 lessons · 80 min expand_more
  • lockOpenAI-compatible vs Anthropic native API format 18 min
  • lockBuilding a provider abstraction layer in PHP 22 min
  • lockAdding Gemini, Grok, and Groq to the dispatcher 20 min
  • lockModule project: Switch providers at runtime from a dropdown 20 min
code
Lab Outcome
One chat interface, four providers. Selecting a model in the UI routes to the correct API without changing a line of core logic.
04 Local AI with Ollama 5 lessons · 90 min expand_more
  • lockInstalling Ollama on Linux and your local machine 18 min
  • lockPulling and running Llama 3.2, Mistral, and Phi-4 16 min
  • lockConnecting Ollama to your PHP dispatcher 20 min
  • lockRunning Ollama as a systemd service 14 min
  • lockModule project: Zero-cost, zero-data-leak chat with Llama 22 min
code
Lab Outcome
Llama 3.2 running on your server, responding to chat requests through the same interface as your cloud providers. No API costs. No data leaving your machine.
05 Document Intelligence 5 lessons · 95 min expand_more
  • lockExtracting text from PDF, DOCX, and CSV in PHP 20 min
  • lockSending documents to the AI API as context 18 min
  • lockSummarise, extract, compare, and draft replies from docs 16 min
  • lockSecure file uploads: MIME validation, size limits, storage 18 min
  • lockModule project: Document Q&A interface — upload and interrogate any file 23 min
code
Lab Outcome
Upload a PDF, ask questions about it in natural language, get AI-powered answers with cited page references — deployed on your server.
06 RAG — Retrieval Augmented Generation 5 lessons · 100 min expand_more
  • lockWhat RAG is and why it matters — hallucination control 18 min
  • lockText chunking strategies: fixed, sentence, and semantic 20 min
  • lockEmbeddings and vector similarity search in PHP 22 min
  • lockBuilding a knowledge base: store, retrieve, and inject context 22 min
  • lockModule project: AI chatbot grounded in your own documents 18 min
code
Lab Outcome
A knowledge base system: add documents, chunk them, store embeddings in MariaDB, and query with cosine similarity. The AI only answers from your data.
07 Prompt Engineering in Code 4 lessons · 75 min expand_more
  • lockSystem prompts, personas, and output format enforcement 18 min
  • lockBuilding a reusable prompt template engine in PHP 22 min
  • lockVariable injection, validation, and preview 16 min
  • lockModule project: Visual prompt builder with live output preview 19 min
code
Lab Outcome
A prompt library with a visual builder — create templates, define variable fields, preview with test data, and fire live API calls.
08 Content Generation Pipeline 4 lessons · 80 min expand_more
  • lockMulti-step generation chains: outline → draft → edit 20 min
  • lockTone and brand voice control via system prompts 16 min
  • lockGenerating structured output: JSON, HTML, Markdown 18 min
  • lockModule project: Blog post generator from a brief 26 min
code
Lab Outcome
Input a brief, select a tone, choose a provider — a multi-step pipeline outputs a formatted blog post stored in your database.
09 Image Generation 4 lessons · 75 min expand_more
  • lockDALL-E 3 via the OpenAI API from PHP 18 min
  • lockStability AI SDXL: style presets and aspect ratios 18 min
  • lockAI prompt enhancement — rewrite user input for better results 16 min
  • lockModule project: Image studio with style picker and gallery 23 min
code
Lab Outcome
A live image studio: enter a prompt, pick a style, AI rewrites it, DALL-E generates the image, it's saved to your server and displayed in a gallery.
10 AI Security — Hardening Your Application 5 lessons · 90 min expand_more
  • lockEncrypting API keys with AES-256-GCM in PHP 18 min
  • lockPrompt injection attacks and how to prevent them 20 min
  • lockInput sanitisation before and after the AI call 16 min
  • lockPOPIA considerations when passing user data to AI 18 min
  • lockModule project: Security audit of your AI application 18 min
code
Lab Outcome
All API keys encrypted at rest, prompt injection attempts blocked, AI outputs sanitised, POPIA data handling documented.
11 Audit Logging & Monitoring AI Calls 4 lessons · 70 min expand_more
  • lockWhat to log: prompt, model, tokens, latency, cost estimate 16 min
  • lockBuilding an immutable audit log table 18 min
  • lockVisualising token usage and cost over time 18 min
  • lockModule project: Admin audit dashboard for all AI calls 18 min
code
Lab Outcome
Every AI call logged to an immutable table — model used, tokens consumed, latency, cost estimate — visible in an admin dashboard.
12 Capstone: Full AI Application 6 lessons · 120 min expand_more
  • lockArchitecture planning: modules, database schema, API structure 20 min
  • lockUser authentication and role-based access for AI features 20 min
  • lockIntegrating all modules: chat, docs, RAG, image gen 24 min
  • lockProduction hardening: rate limiting, error handling, logging 18 min
  • lockDeployment to your VPS with database migrations 20 min
  • lockCapstone showcase and peer review 18 min
code
Capstone Outcome
A fully functional multi-provider AI application — chat, document intelligence, knowledge base, image generation — live on your own VPS, hardened, logged, and ready for users.

Requirements

  • check_circleCompletion of Tech Stack & Hosting, or equivalent server experience
  • check_circleA working VPS with Apache/Nginx, PHP 8.1+, and MariaDB
  • check_circleBasic PHP knowledge — variables, functions, arrays, forms
  • check_circleAPI keys for at least one provider (Claude or OpenAI — we help you get them)
  • cancelNo framework knowledge required — pure PHP throughout
  • cancelNo prior AI development experience needed

Who this course is for

  • personPHP developers ready to add real AI capabilities to their applications
  • personAnyone who completed Tech Stack & Hosting and wants to build on it
  • personDevelopers wanting to build and sell AI-powered tools
  • personTechnical founders building AI-first products who want full ownership of their stack

Your Instructor

VM
VoxMachina Team
AI Application Engineers · VoxMachina
Everything in this course is code we've shipped. VoxGen, ContentFlow, PaymentFlow — every one uses the exact patterns you'll build here: multi-provider dispatch, streaming SSE, RAG, encrypted keys, full audit logs. You're learning from working production code.
6AI Apps in Production
5+AI Providers Integrated
RAGProduction RAG Systems

Frequently Asked Questions

helpDo I need to buy API access before starting?expand_more
You need at least one API key — Anthropic or OpenAI. Both have free trial credits sufficient for Modules 1–4. In Module 4 we also set up Ollama for completely free local inference, so you can work through most of the course with zero API costs. We walk through getting API keys in Module 1.
helpDo I need to take Tech Stack & Hosting first?expand_more
Recommended but not strictly required. If you already have a working VPS with Apache, PHP 8.1+, and MariaDB configured, you can jump straight here. If you're not sure whether your server is ready, attend the first session — we'll check it together.
helpIs this course in JavaScript or PHP?expand_more
Primarily PHP for the backend — all API calls, database logic, security, and server-side processing. JavaScript is used for the frontend: SSE streaming, fetch() calls, and dynamic UI updates. No React or Node.js — this is vanilla JS talking to PHP endpoints.
helpWhat do I actually build in the capstone?expand_more
A full AI workspace application: multi-provider chat with streaming, document upload and Q&A, a knowledge base with RAG, a content generation tool with tone control, and an image generator. Everything deployed on your own server — essentially a stripped-down version of VoxGen.

Continue Your Learning

The final step after AI Development:

rocket_launch
Full Stack Bootcamp
R14 999 · All courses + Capstone
apps
Explore Our AI Apps
See what you'll be able to build