๐ Deskripsi Kursus
Pelatihan tingkat lanjut untuk mendalami Generative AI (ChatGPT, Gemini, Claude, LLaMA, dll.) dengan fokus pada teknik prompt engineering lanjutan, fine-tuning model, integrasi API, serta pengembangan custom tool berbasis LLM. Kursus ini dirancang agar peserta mampu membangun workflow AI yang kompleks, mengintegrasikan data internal perusahaan, serta menciptakan solusi AI yang etis, aman, dan bernilai bisnis nyata.
๐ฏ Tujuan Pembelajaran
- Memahami arsitektur LLM, kapabilitas mutakhir, keterbatasan, serta prinsip etika & governance.
- Merancang prompt kompleks: few-shot, chain-of-thought, tree-of-thought, persona, multi-turn dialog.
- Mengintegrasikan LLM melalui API dengan Python, LangChain, Hugging Face, dan platform serupa.
- Menerapkan fine-tuning model (LoRA, QLoRA, PEFT) untuk kebutuhan domain khusus.
- Mengembangkan sistem Retrieval-Augmented Generation (RAG) menggunakan vector database & embedding.
- Merancang custom AI tools & micro-workflows untuk divisi bisnis (legal, HR, marketing, finance, dsb).
- Menyusun Prompt Library, template, dan SOP AI lintas fungsi di perusahaan.
๐๏ธ Struktur Kursus (6 Sesi โ 3 Jam/Sesi)
Sesi | Topik | Fokus Utama |
---|---|---|
1 | Advanced LLM & Governance | Transformer, attention, model comparison, enterprise risk & governance |
2 | Advanced Prompt Engineering | Few-shot, chain-of-thought, tree-of-thought, persona, multi-modal prompting |
3 | API Integration & Light Coding | Python scripting, LangChain, Hugging Face, building bots & workflows |
4 | Model Fine-Tuning | LoRA, QLoRA, RLHF, data prep, evaluation metrics, hands-on fine-tuning |
5 | Retrieval & Multimodal AI | RAG pipelines, vector DBs, embeddings, multimodal use cases |
6 | Implementation & Presentation | Team workshop: build micro-workflows, prompt library, final presentation |
๐งช Contoh Proyek/Latihan
- Fine-tune a model on internal FAQs (HR/Customer Support).
- Build a RAG Q&A bot using internal documents.
- Create a Slack/Sheets AI assistant with API integration.
- Develop multi-step summarizer (meeting notes โ action items).
- Design prompt packs for Marketing, Legal, Finance divisions.
- Build an AI governance โbias checkerโ for generated outputs.
๐ Tools & Software Requirements
- ChatGPT / Gemini / Claude / LLaMA
- Python 3, Jupyter/Colab, VSCode
- Hugging Face Transformers, LangChain, LlamaIndex
- Vector Database (FAISS, Pinecone, Weaviate)
- Google Docs / Sheets / Notion (for workflow integration & docs)
๐ Referensi
- OpenAI Documentation & API Reference
- Hugging Face Transformers & PEFT Docs
- LangChain Documentation
- Google Cloud Prompt Engineering Guide
- Academic papers (LoRA, QLoRA, RAG, RLHF)
๐งฎ Skema Penilaian
Component | Weight |
---|---|
Kehadiran & Partisipasi | 20% |
Tugas / Latihan | 20% |
Mini-Project | 30% |
Presentasi Akhir | 30% |
๐ Prasyarat
- Telah menyelesaikan kursus Beginner/Intermediate atau memiliki pengalaman setara.
- Familiar dengan Python dasar (variabel, fungsi, API calls).
- Literasi digital & pemahaman proses bisnis.