1Z0-1127 Study Guide — How to Pass Oracle Cloud Infrastructure Generative AI Professional

Week 1–2

LLM Fundamentals: transformer architecture, tokenisation, embeddings, attention mechanism, pre-training vs fine-tuning

Tip: The OCI GenAI exam goes deeper on LLM internals than most foundational AI certs. Know how attention works conceptually, what embeddings represent (dense vector representations of semantic meaning), and the difference between encoder-only (BERT — understanding), decoder-only (GPT — generation), and encoder-decoder (T5 — translation/summarisation) architectures.

Week 3–4

OCI Generative AI Service: chat models (Cohere Command, Meta Llama), embedding models, dedicated AI clusters, fine-tuning on OCI

Tip: Know the OCI GenAI service models: Cohere Command R/R+ for chat and RAG; Cohere Embed for generating embeddings; Meta Llama for open-source generation. Know OCI dedicated AI clusters (private GPU infrastructure) vs shared infrastructure, and when each is appropriate for enterprise workloads.

Week 5–6

Prompt Engineering & RAG: zero-shot, few-shot, chain-of-thought, RAG architecture, vector databases, OCI OpenSearch

Tip: RAG is a major exam topic. Know the full RAG pipeline: document ingestion → chunking → embedding → vector store (OCI OpenSearch with k-NN) → retrieval → prompt augmentation → LLM generation. Know why RAG reduces hallucinations and how chunk size affects retrieval quality.

Week 7–8

LangChain on OCI: chains, agents, memory, tools, LangChain OCI integration, building production GenAI applications

Tip: LangChain is heavily tested. Know the core abstractions: LLMs/ChatModels (model interface), Chains (sequences of calls), Agents (LLM + tools for dynamic reasoning), Memory (conversation history), and Vector Stores (retrieval). Know how to integrate OCI GenAI models as LangChain LLMs using the OCI LangChain integration.