Hands-on AI Engineering
Learn AI Engineering
by shipping it.
Theory you read, code you run. Every lesson pairs writing with a live sandbox that calls real AI models — securely proxied, so no API keys ever touch the browser.
Live sandboxes
The sandbox is the lesson.
No copy-pasting into a scratch project. You edit real code in the browser, it hits our secure proxy, and a live model answers — token by token.
- Keys stay server-side — the proxy injects them.
- Streaming responses, just like production.
- Validate your output to unlock the next lesson.
const res = await fetch(PROXY_URL + "/api/sandbox-proxy", {
method: "POST",
headers: { authorization: `Bearer ${TOKEN}` },
body: JSON.stringify({
provider: "openai",
model: "gpt-4o-mini",
mode: "stream",
messages: [{ role: "user", content: "What is a token?" }],
}),
});
// → streams the answer, token by tokenThe curriculum
Six modules, zero fluff
From your first API call to a monitored, evaluated agent in production.
LLMs as a Service
The raw API, tokens, context windows, params, and streaming.
Prompting & Structured Outputs
Zod-validated JSON you can actually trust in production.
AI Frameworks
The Vercel AI SDK — and when not to reach for LangChain.
Memory & Context (RAG)
Embeddings, chunking, vector search, cosine similarity.
Action & Orchestration
Tool calling and ReAct agents that take real actions.
Production (LLMOps light)
Retries, rate limits, observability, and evals.
Skills & Context
Augment skills, context engineering, and prompt management.
Model Context Protocol (MCP)
Standardized tools, servers, and integrations for AI models.
Learning that runs
No slides, no throwaway snippets. Each step is a working sandbox.
Read & write
Short theory, then you edit real code in an in-browser editor.
Run for real
Your code hits our secure proxy and calls a live AI model.
Validate to advance
We check your output and unlock the next lesson.
Ready to build with AI?
Start with Module 1 and make your first real model call in the next five minutes.
Open the first lesson