Skip to content

OCI Generative AI Professional (1Z0-1127-25)

Building generative AI applications on Oracle Cloud Infrastructure: LLM fundamentals, the OCI Generative AI service, and RAG applications with LangChain and vector search.

Together with OCI AI Foundations, this closes the gap where the repo carried five Oracle certifications and no Oracle AI ones.

Exam Details

  • Exam Code: 1Z0-1127-25
  • Duration: 90 minutes
  • Questions: 60, multiple choice
  • Passing Score: 68%
  • Cost: USD 245 list; free certification windows appear periodically
  • Prerequisites: None formal; AI Foundations and Python familiarity recommended

Full detail in the fact sheet.

Notes

Notes Covers
01 LLM fundamentals Transformers, tokens, decoding parameters, prompting, customization spectrum
02 The OCI Generative AI service Models, on-demand versus dedicated clusters, fine-tuning, security
03 Building LLM applications RAG, chunking, vector search, LangChain, evaluation, deployment

The three decisions the exam keeps testing

1. How to customize a model. Prompting, then RAG, then fine-tuning, in that order of preference. Knowledge problems are RAG problems; behavior and format problems are fine-tuning problems. An answer that reaches for fine-tuning to give a model access to company documents is wrong.

2. On-demand or dedicated cluster. On-demand for variable, low-volume, or experimental workloads, billed per use. Dedicated AI clusters for predictable throughput, isolation, and any custom fine-tuned model, billed for the reserved capacity. Fine-tuning itself requires a cluster.

3. Where retrieval quality comes from. Chunking strategy, embedding model choice, similarity metric, and how many chunks you retrieve. Most "the answers are bad" scenarios are retrieval problems rather than model problems.

Hands-on

An OCI free tier account plus the Generative AI playground covers most of this. Worth doing:

  • Run the same prompt at temperature 0 and temperature 1 and compare determinism
  • Compare zero-shot, few-shot, and chain-of-thought on a reasoning task
  • Generate embeddings and compute cosine similarity between related and unrelated sentences
  • Build a small RAG pipeline: chunk a document set, embed, store, retrieve, and answer
  • Change the chunk size and observe the effect on answer quality
  • Set up an OCI Generative AI Agent over a document store

Study resources