Oracle OCI AI Foundations Study Guide & Cheat Sheet

Oracle OCI AI Foundations Associate (1Z0-1122-25) Study Guide

Exam domains

Who it’s for: Anyone wanting foundational AI/ML knowledge plus familiarity with Oracle Cloud Infrastructure AI services. It is an accessible, free entry-level certification suited to students, business users, and technical professionals beginning their AI journey on OCI.

Study & test-day tips

• OCI AI Services carries the most weight (20%), so prioritize knowing what OCI Language, Vision, Speech, Document Understanding, and Anomaly Detection each do and when to choose one over another.
• Be able to clearly separate AI vs. machine learning vs. deep learning vs. generative AI, since many questions test whether you can place a concept at the right level.
• Memorize the three learning paradigms (supervised, unsupervised, reinforcement) and map each to its typical tasks: regression and classification are supervised, clustering is unsupervised.
• For deep learning, connect each architecture to its strength: CNNs for images, RNNs/LSTMs for sequential data, and transformers for language and modern LLMs.
• Know the LLM workflow vocabulary cold: prompt engineering, fine-tuning, instruction tuning, embeddings, vector search, and retrieval-augmented generation (RAG).
• Understand that embeddings are numeric vector representations and that vector search powers semantic retrieval, which is the foundation of RAG.
• Distinguish OCI Generative AI (managed LLM service) from OCI Generative AI Agents (agentic apps with RAG over your data) and from OCI Data Science (build, train, deploy custom models).
• Remember the Oracle Database 23ai AI features: AI Vector Search for similarity search inside the database and Select AI for natural-language querying of your data.
• Take the free official Oracle training course before the exam; it is aligned directly to the objectives and the questions track its terminology closely.
• You only need 60% across 30 questions in 60 minutes, so pace yourself at roughly two minutes per question, flag uncertain items, and answer every question since there is no penalty for guessing.

Cheat sheet

AI Concept Hierarchy

• Artificial Intelligence (AI): broad field of machines performing tasks that mimic human intelligence.
• Machine Learning (ML): subset of AI where systems learn patterns from data instead of explicit rules.
• Deep Learning (DL): subset of ML using multi-layer neural networks to learn complex representations.
• Generative AI: models that create new content (text, images, code) from learned patterns.
• Responsible AI: building systems that are fair, transparent, accountable, and respect privacy.

Machine Learning Paradigms & Tasks

• Supervised learning: trains on labeled data; used for regression and classification.
• Unsupervised learning: finds structure in unlabeled data; used for clustering.
• Reinforcement learning: an agent learns by trial and error via rewards and penalties.
• Regression: predicts a continuous numeric value (e.g., price, temperature).
• Classification: predicts a discrete category or label (e.g., spam vs. not spam).
• Clustering: groups similar data points without predefined labels.

Deep Learning Architectures

• Neural network: layers of interconnected nodes (neurons) that learn weights from data.
• CNN (Convolutional Neural Network): excels at images and computer vision tasks.
• RNN / LSTM: handles sequential data such as text and time series; LSTMs retain longer context.
• Transformer: attention-based architecture that underpins modern LLMs.
• Training vs. inference: training learns model parameters; inference applies the model to new data.

Generative AI & LLM Toolkit

• Large Language Model (LLM): transformer model trained on vast text to generate and understand language.
• Prompt engineering: crafting inputs to guide an LLM toward the desired output.
• Fine-tuning: further training a pretrained model on domain-specific data.
• Instruction tuning: fine-tuning a model to follow natural-language instructions.
• Embeddings + vector search: numeric vectors enabling semantic similarity search.
• RAG (Retrieval-Augmented Generation): retrieves relevant data to ground an LLM's responses.

OCI AI Portfolio & Infrastructure

• OCI AI Services: ready-to-use pretrained (and customizable) services via API.
• OCI Data Science: managed platform to build, train, deploy, and manage ML models.
• OCI Generative AI: managed service offering pretrained and fine-tunable foundation models.
• OCI Generative AI Agents: build agentic apps that use RAG over your enterprise data.
• AI infrastructure: GPU-based compute, including NVIDIA GPUs, for training and inference.

OCI AI Services & Database 23ai

• OCI Language: text analysis such as sentiment, entities, language detection, and translation.
• OCI Vision: image analysis including object detection, classification, and text/OCR extraction.
• OCI Speech: converts spoken audio into text (speech-to-text transcription).
• OCI Document Understanding: extracts text, tables, and key-value data from documents.
• OCI Anomaly Detection: identifies unusual patterns in time-series and multivariate data.
• Oracle Database 23ai: AI Vector Search for similarity search and Select AI for natural-language queries.

Glossary

Artificial Intelligence (AI)

The broad field of building machines and software that perform tasks normally requiring human intelligence, such as reasoning, perception, and language.

Machine Learning (ML)

A subset of AI in which systems learn patterns from data to make predictions or decisions without being explicitly programmed with rules.

Deep Learning (DL)

A subset of machine learning that uses multi-layer neural networks to automatically learn complex representations from large amounts of data.

Supervised Learning

A learning approach that trains models on labeled examples so they can predict outputs for new inputs; used for regression and classification.

Unsupervised Learning

A learning approach that discovers structure or groupings in unlabeled data, commonly used for clustering and dimensionality reduction.

Reinforcement Learning

A learning approach where an agent learns optimal actions by interacting with an environment and receiving rewards or penalties.

Regression

A supervised learning task that predicts a continuous numeric value, such as a price or temperature.

Classification

A supervised learning task that assigns inputs to discrete categories or labels, such as spam versus not spam.

Clustering

An unsupervised learning task that groups similar data points together without using predefined labels.

Neural Network

A model made of layers of interconnected nodes (neurons) whose weighted connections are learned from data.

Convolutional Neural Network (CNN)

A neural network architecture specialized for grid-like data such as images, widely used in computer vision tasks.

Recurrent Neural Network (RNN)

A neural network designed to process sequential data by maintaining state across steps in the sequence.

Long Short-Term Memory (LSTM)

A type of recurrent neural network that uses gating to retain information over longer sequences and mitigate vanishing gradients.

Transformer

An attention-based neural network architecture that processes sequences in parallel and underpins modern large language models.

Large Language Model (LLM)

A transformer-based model trained on massive text corpora to understand and generate human-like language.

Generative AI

AI that creates new content such as text, images, audio, or code based on patterns learned from training data.

Prompt Engineering

The practice of designing and refining input prompts to guide a language model toward desired and accurate outputs.

Fine-Tuning

The process of further training a pretrained model on domain- or task-specific data to improve its performance.

Instruction Tuning

Fine-tuning a model on instruction-and-response examples so it more reliably follows natural-language instructions.

Embeddings

Numeric vector representations of text or other data that capture semantic meaning for comparison and search.

Vector Search

Searching for items by similarity of their embedding vectors, enabling semantic rather than keyword matching.

Retrieval-Augmented Generation (RAG)

A technique that retrieves relevant external data and supplies it to an LLM to ground and improve its responses.

Responsible AI

Principles and practices for building AI that is fair, transparent, accountable, secure, and respectful of privacy.

Inference

The stage where a trained model is applied to new input data to produce predictions or generated output.

OCI Generative AI

A fully managed OCI service offering access to pretrained and fine-tunable foundation models through an API.

OCI Generative AI Agents

An OCI offering for building agentic, LLM-powered applications that use retrieval-augmented generation over enterprise data.

OCI Data Science

A managed OCI platform for data scientists to build, train, deploy, and manage machine learning models.

OCI Language

An OCI AI service for text analysis, including sentiment analysis, entity and key-phrase extraction, language detection, and translation.

OCI Vision

An OCI AI service that analyzes images for object detection, image classification, and text extraction (OCR).

OCI Speech

An OCI AI service that transcribes spoken audio into text using automatic speech recognition.

OCI Document Understanding

An OCI AI service that extracts text, tables, and key-value information from documents such as invoices and forms.

OCI Anomaly Detection

An OCI AI service that identifies unusual patterns or outliers in time-series and multivariate data.

Oracle Database 23ai AI Vector Search

A capability in Oracle Database 23ai that stores embeddings and performs similarity (vector) search alongside relational data.

Select AI

An Oracle Database feature that lets users query their data using natural language, which is translated into SQL.