AI Engineer Roadmap — Currículo de Autoestudo
satyajit-mohapatra/ai-engineer-roadmap
Roteiro completo de IA: contexto, prompts, RAG, agents e modelos
Ver no GitHubCurrículo estruturado e interativo para aprender engenharia de IA seguindo o roadmap.sh/ai-engineer, com 29 lições em HTML divididas em duas fases. Integra-se com Claude Code via skill /teach para ensino personalizado e rastreamento de progresso contínuo.
Instalação
npx convez-ai satyajit-mohapatra/ai-engineer-roadmap1 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
Casos de uso
- Aprender engenharia de IA desde o zero de forma estruturada
- Dominar hooks, skills e MCP no ecossistema Claude Code
- Estudar LLMs, prompts, RAG, embeddings e fine-tuning
- Rastrear progresso com notas vivas e quizzes integrados
Veredito

Recurso educacional bem estruturado e autoexplicativo para quem quer dominar engenharia de IA no contexto Claude. A integração com /teach skill do Claude Code torna a experiência verdadeiramente interativa, mas exige disciplina e acesso à ferramenta para máximo aproveitamento.
README
A tradução para o português deste README ainda está na fila.
AI Engineering — Self-Study Workspace
A self-paced curriculum following the roadmap.sh/ai-engineer
path, built and taught interactively with Claude Code's /teach skill.
Structure
| Path | What it is |
|---|---|
MISSION.md | The learning goal, success criteria, and scope/constraints. Read this first. |
RESOURCES.md | Reference material and links used across lessons. |
NOTES.md | Living notes: user profile, preferences, curriculum status, session log. |
reference/curriculum.html | Index of all lessons, grouped by phase. Open this to navigate. |
reference/glossary.html | Glossary of terms used across lessons. |
lessons/ | Self-contained HTML lessons (0001–0029), each with a quiz and source links. |
learning-records/ | Personal session reflections — git-ignored, kept local only. |
Curriculum
29 lessons in two phases:
- Phase 1 (1–13) — Context & harness engineering for Claude Code: hooks, skills, CLAUDE.md, memory, MCP, context window management, subagents, loop engineering.
- Phase 2 (14–29) — The rest of the AI Engineer roadmap: how LLMs work, choosing models, inference parameters, provider APIs, prompt engineering, AI safety, open-source models, embeddings, vector databases, RAG, agents & function calling, multimodal AI, the model provider landscape, fine-tuning, and AI dev tools.
Open reference/curriculum.html in a browser to browse and
jump between lessons.
How to use this
- Browse lessons by opening any file in
lessons/directly in a browser, or start from the curriculum index. - Take a lesson: read through the material, try the embedded examples, and answer the two quiz questions at the end of each lesson.
- Continue learning with Claude Code: open this workspace in Claude Code and run
the
/teachskill. It readsMISSION.mdandNOTES.mdto pick up where you left off, generates new lessons, and updates the curriculum index as you progress. - Track progress:
NOTES.mdrecords curriculum status and session history so a new session can resume seamlessly.
Notes
- Lessons are self-contained HTML files (dark theme, no build step) — just open them in a browser.
learning-records/is excluded from version control (see.gitignore) since it contains personal reflections, not shareable lesson content.