convez.ai
AgenteDesenvolvimento WebIniciante
ruhaankumar2013-debug

Currículo Prático de Engenharia LLM

ruhaankumar2013-debug/Practical-AI-engineering

Do 'Hello, GPT' até sistemas de IA em produção — currículo prático e comunitário

Ver no GitHub
BomPluginScore · de 0 a 100

Currículo open-source completo que ensina desenvolvimento com Large Language Models, desde prompting básico até implantação de sistemas em produção. Cobre engenharia de prompts, RAG, fine-tuning, agents e DevOps, com código executável e projetos portfólio em cada fase. Mantido pela comunidade de praticantes, framework-agnóstico e estruturado em 6 fases do iniciante ao avançado.

Instalação

$npx convez-ai ruhaankumar2013-debug/Practical-AI-engineering

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
16
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Aprender LLMs do zero até produção com código executável
  • Preparar portfólio técnico com projetos capstone em cada fase
  • Atualizar conhecimento em arquiteturas de IA sem framework específico

Veredito

Currículo bem estruturado e hands-on que cobre genuinamente a jornada completa do LLM. Força real é o escopo 360° (prompting→RAG→fine-tuning→agents→produção) com foco em código real, não teoria. Limitação: qualidade depende da manutenção comunitária contínua e da cobertura prática dos tópicos avançados.

README

A tradução para o português deste README ainda está na fila.

🧠 LLM Dev Curriculum

The open-source, community-driven roadmap to building with Large Language Models —

from your first "Hello, GPT" to shipping production AI systems.

Stars Contributors License: MIT PRs Welcome Last Updated

⭐ If this helped you, star it — it helps others find it too.

🚀 Start Learning · 📚 All Modules · 🏗️ Projects · 🤝 Contribute · 💬 Community


🌟 Why This Curriculum?

There are hundreds of LLM tutorials out there. Most are either:

  • 🔴 Too shallow — "just call the API and you're done"
  • 🔴 Too academic — dense math, no runnable code
  • 🔴 Too opinionated — locked into one framework
  • 🔴 Outdated — written in 2023 and never touched since

This curriculum is different. It's:

✅What you get
Beginner → ProductionStart from zero ML knowledge. End with a deployed, monitored AI system.
Framework-agnosticOpenAI, Anthropic, HuggingFace, open-weight models — you learn all of them.
Hands-on firstEvery concept has code. Every phase has a capstone project you can put on your portfolio.
Community-maintainedUpdated by practitioners, not just educators. PRs welcome.
Clearly leveledEvery module is tagged 🟢 Beginner · 🟡 Intermediate · 🔴 Advanced — no surprises.

🗺️ The Big Picture

YOU ARE HERE
     │
     ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 0 · Foundations                    🟢 No prerequisites │
│  What LLMs are, how transformers work, tokens & embeddings   │
└───────────────────────────┬──────────────────────────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 1 · Prompting                           🟢 Beginner   │
│  Prompt engineering, CoT, few-shot, system prompts           │
└───────────────────────────┬──────────────────────────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 2 · APIs & Integrations                 🟢 Beginner   │
│  Call LLMs from code. Build your first real app.             │
└───────────┬───────────────┴───────────────┬──────────────────┘
            │                               │
            ▼                               ▼
┌───────────────────────┐       ┌───────────────────────────┐
│  PHASE 3 · RAG  🟡    │       │  PHASE 4 · Fine-Tuning 🟡 │
│  Retrieval-Augmented  │       │  LoRA, QLoRA, RLHF,       │
│  Generation, vector   │       │  dataset prep             │
│  DBs, chunking        │       └───────────┬───────────────┘
└───────────┬───────────┘                   │
            └───────────────┬───────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 5 · Agents                         🟡 Intermediate    │
│  Tool use, ReAct, memory, multi-agent systems                │
└───────────────────────────┬──────────────────────────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 6 · Evaluation                     🟡 Intermediate    │
│  Evals, benchmarks, LLM-as-judge, eval pipelines            │
└───────────────────────────┬──────────────────────────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 7 · Production                          🔴 Advanced   │
│  Deploy, scale, monitor, optimize, guardrails                │
└───────────────────────────┬──────────────────────────────────┘
                            │
                            ▼
┌──────────────────────────────────────────────────────────────┐
│  PHASE 8 · Advanced Topics                     🔴 Advanced   │
│  Train from scratch, multimodal, MoE, research papers        │
└──────────────────────────────────────────────────────────────┘

🏁 Quick Start

Not sure where to begin? Pick your entry point:

🟢 "I know nothing about LLMs or ML"

Start → 00-foundations_what_are_llms.md

🟢 "I know ML / data science, new to LLMs"

Start → 02-prompt_engineering.md

🟢 "I'm a developer, just want to build an app NOW"

Start → 01-apis_quickstart.md

🟡 "I want to connect LLMs to my own data"

Start → 03-rag_pipeline.md

🟡 "I want to customize / fine-tune a model"

Start → 04-finetuning_lora.md

🔴 "I need to ship this to production"

Start → 06-safety_guardrails.md

📚 Modules

Legend: 📖 Guide · 🧪 Notebook · 💻 Code · 🏗️ Project · 🟢 Beginner · 🟡 Intermediate · 🔴 Advanced


🟢 Phase 0 — Foundations

No prerequisites. If you're brand new, start here.

#ModuleFormatEst. Time
0What Are LLMs?📖20 min

🟢 Phase 1 — Prompting

Learn to communicate with LLMs effectively. The highest ROI skill in this entire curriculum.

#ModuleFormatEst. Time
1Prompt Engineering Techniques📖 + examples1 hr

🟢 Phase 2 — APIs & Integrations

Call LLMs from your code. Build your first real application.

#ModuleFormatEst. Time
2OpenAI API Quickstart📖 + 💻1 hr

🟡 Phase 3 — RAG (Retrieval-Augmented Generation)

Give your LLM a memory. Connect it to your own documents and data.

#ModuleFormatEst. Time
3Building a RAG Pipeline End-to-End📖 + 🧪2 hr

🟡 Phase 4 — Fine-Tuning

When prompting isn't enough — adapt a model to your exact task.

#ModuleFormatEst. Time
4LoRA and QLoRA: Efficient Fine-Tuning📖 + 🧪3 hr

🔴 Phase 5 — Production

Ship it. Scale it. Keep it working. Don't go broke.

#ModuleFormatEst. Time
5Safety, Guardrails & Content Filtering📖 + 💻1.5 hr

📖 Additional Resources

ResourceLink
📚 Glossary05-glossary.md

🏗️ Projects Summary

Each phase ends with a capstone project you can actually put on your portfolio.

PhaseProjectSkills Demonstrated
0FoundationsLLM basics
1Prompt PlaygroundPrompt engineering
2CLI ChatbotAPI calls, conversation history
3Chat with Your DocsRAG, vector DB, embeddings
4Custom Fine-TuneDataset prep, LoRA, model training
6Safety ImplementationGuardrails, content filtering

🗂️ Repository Structure

Practical-AI-engineering/
├── 00-foundations_what_are_llms.md
├── 01-apis_quickstart.md
├── 02-prompt_engineering.md
├── 03-rag_pipeline.md
├── 04-finetuning_lora.md
├── 05-glossary.md
├── 06-safety_guardrails.md
├── README.md                 ← you are here
├── contributing.md
├── roadmap.md
└── LICENSE

🔖 Format Legend

IconMeaning
📖Written guide (Markdown)
🧪Jupyter Notebook (runnable)
💻Standalone code / project
🏗️Capstone project
🟢Beginner — no prior ML knowledge needed
🟡Intermediate — comfortable with Python and APIs
🔴Advanced — ML background helpful

🤝 Contributing

This curriculum is community-powered. All contributions welcome — big or small.

  • Typo / small fix → open a PR directly
  • Improve an explanation → open a PR with a brief note on what was unclear
  • Add a new module → open an issue first so we can align on scope
  • Translate to your language → see the Translations section

Please read contributing.md before submitting. Be kind.


🌍 Translations

LanguageStatusMaintainer
🇬🇧 English✅ CompleteCore Team
🇮🇳 Hindi🚧 In ProgressContribute →
🇪🇸 Spanish🚧 In ProgressContribute →
🇵🇹 Portuguese🙋 NeededContribute →
🇨🇳 Chinese (Simplified)🙋 NeededContribute →
🇯🇵 Japanese🙋 NeededContribute →

💬 Community

  • GitHub Discussions — questions, study groups, project showcases
  • Issues — bug reports, content errors, suggestions

📜 License

MIT — free to use, share, fork, and build on.
Attribution appreciated but not required.


Built with ❤️ by the community · Contribute · Star it ⭐

This curriculum is updated regularly. Watch the repo to stay notified.

Ver o original no GitHub

Bom para

Relacionados