OCI Generative AI Professional
Overview
This document covers the Oracle Cloud Infrastructure 2025 Generative AI Professional certification (1Z0-1127), focusing on large language models, embeddings, RAG, and OCI AI services.
Core Concepts
Large Language Models on OCI
OCI Generative AI service provides access to foundation models including Cohere Command, Meta Llama, and Oracle's own models for text generation, summarisation, and embeddings.
Key Terms
| Term | Definition |
|---|---|
| Token | Basic unit of text (~4 chars); models have token limits |
| Embedding | Dense vector representation of text for semantic search |
| RAG | Retrieval-Augmented Generation, inject retrieved docs into LLM context |
| Fine-tuning | Adapt a foundation model to a specific domain with labelled data |
| Prompt | Input instruction/context given to the model |
OCI AI Services
- OCI Generative AI, managed LLM inference and fine-tuning.
- OCI AI Language, NLP: sentiment, key phrases, NER, translation.
- OCI Vision, image classification, object detection, document AI.
- OCI Speech, transcription and text-to-speech.
- OCI Anomaly Detection, time-series anomaly detection.
RAG Architecture
User Query → Embedding Model → Vector Search (OCI OpenSearch / 23ai)
→ Retrieved Chunks → LLM Context → Generated Response
Cheat Sheet
OCI GenAI: Cohere | Llama | Text generation | Summarisation | Embeddings
RAG pipeline: Embed → Store → Retrieve → Generate
Fine-tuning: T-Few | Vanilla, domain adaptation