Command Palette

Search for a command to run...

Hectal

LLM Engineering · from next-token prediction to production AI features

Build real products on LLMs.
In Python, Java and JavaScript.

You need the basics of one programming language and nothing about AI. Every topic starts from why the thing exists, builds the simplest version by hand, shows where it breaks, and only then reaches for the real API or library. Every key example runs on the page in Python, Java and JavaScript and prints the same output, using small fake models so runs are free and repeatable. Real-model labs send your own requests to a free hosted model and show the real reply, tokens, timing and limits. Every phase ends with a playbook of recipes you can copy into real work.

Start from zero The full roadmap
Phases
15
Topics
128
Hours
46
Diagrams
129
Exercises
737

By the end you can

What this course makes you able to do

  • Explain how LLMs work as an engineer: tokens, embeddings, attention, context windows, sampling and why models hallucinate
  • Call models over raw HTTP and through SDKs in Python, Java and JavaScript, with timeouts, error handling and keys kept safe
  • Get reliable structured output, stream replies, and control cost, latency and rate limits with caching, retries and fallbacks
  • Build tool calling and a working RAG pipeline by hand, then with real libraries
  • Test LLM features with evaluation sets, defend them against prompt injection and data leaks, and monitor them in production
  • Design, deploy and choose models for a production LLM service, and answer LLM system design interview questions

Beginner → Advanced

The curriculum, by level

Beginner
  1. 0Start Here: What an LLM Is and Why Engineering Matters7 topics · ~2 h
  2. 1How LLMs Work (for Engineers)10 topics · ~3 h
  3. 2Calling a Model9 topics · ~3 h
  4. 3Prompting as Engineering9 topics · ~3 h
Intermediate
  1. 4Structured Output8 topics · ~3 h
  2. 5Streaming and UX8 topics · ~3 h
  3. 6Cost, Latency and Limits11 topics · ~4 h
  4. 7Tools and Function Calling9 topics · ~3 h
  5. 8Retrieval Basics (RAG)8 topics · ~3 h
Advanced
  1. 9Evaluation and Testing8 topics · ~3 h
  2. 10Safety and Security (Defensive)9 topics · ~3 h
  3. 11Observability and Operations8 topics · ~3 h
  4. 12Building a Production LLM Service9 topics · ~3 h
  5. 13Choosing and Adapting Models8 topics · ~3 h
  6. 14Interviews and Capstone7 topics · ~3 h

Where should I start?

Pick a path for your goal

New to AI

You can code a little and have never built with an LLM. Start at the first topic: every AI term is explained.

  1. 10 · Start Here: What an LLM Is and Why Engineering Matters
  2. 21 · How LLMs Work (for Engineers)
  3. 32 · Calling a Model
  4. 43 · Prompting as Engineering
  5. 54 · Structured Output

Shipping an LLM feature

You need a feature that works in production: reliable output, sane cost, tools and your own data.

  1. 12 · Calling a Model
  2. 24 · Structured Output
  3. 35 · Streaming and UX
  4. 46 · Cost, Latency and Limits
  5. 57 · Tools and Function Calling
  6. 68 · Retrieval Basics (RAG)

Senior and interviews

You build with LLMs already and want evaluation, security, operations and system design depth.

  1. 19 · Evaluation and Testing
  2. 210 · Safety and Security (Defensive)
  3. 311 · Observability and Operations
  4. 412 · Building a Production LLM Service
  5. 513 · Choosing and Adapting Models
  6. 614 · Interviews and Capstone