How to Pass Oracle OCI AI Foundations (1Z0-1122-25) in 2026: Complete Study Guide

# How to Pass Oracle OCI AI Foundations (1Z0-1122-25) in 2026: Complete Study Guide Artificial intelligence and machine learning are reshaping every industry, and Oracle has built a deep portfolio of AI capabilities directly into Oracle Cloud Infrastructure. The OCI AI Foundations exam (1Z0-1122-25) validates that you understand both the theoretical foundations of AI/ML and the practical services Oracle provides to build AI-powered applications on OCI. This guide walks you through everything you need: exam format, domain breakdown, the most important concepts to master, and a realistic 4-week study plan. --- ## Exam Format at a Glance | Detail | Value | |---|---| | Exam code | 1Z0-1122-25 | | Price | $245 USD | | Number of questions | 60 questions | | Time limit | 90 minutes | | Passing score | 68% (approximately 41/60) | | Difficulty | Easy | | Format | Multiple choice, multiple select | | Delivery | Pearson VUE (online or test center) | The exam is tagged as "easy" difficulty, which reflects the foundational nature of the content. You are not expected to write code or configure OCI services from scratch — you need to understand what each service does, when to use it, and how the pieces fit together in an AI/ML workflow. --- ## Domain Breakdown The exam covers four domains. Oracle does not publish the exact percentage weighting publicly, but based on exam structure and the number of questions (60 total), here is the approximate distribution: | Domain | Topic | Approx. Weight | |---|---|---| | Domain 1 | AI and ML Fundamentals | 25% (~15 questions) | | Domain 2 | OCI AI Services | 30% (~18 questions) | | Domain 3 | Generative AI and Large Language Models | 25% (~15 questions) | | Domain 4 | AI Infrastructure and Data Science | 20% (~12 questions) | Domains 2 and 3 together account for about 55% of the exam. That is where most candidates should focus their preparation time. --- ## Domain 1: AI and ML Fundamentals This domain tests conceptual understanding of machine learning. You do not need to implement algorithms, but you must know what each type of learning does and when it is appropriate. ### Machine Learning Paradigms **Supervised learning** trains a model on labeled data — input-output pairs. The model learns to predict outputs for new inputs. Use cases: classification (spam detection, image labeling), regression (price prediction, demand forecasting). **Unsupervised learning** finds patterns in unlabeled data. The algorithm discovers structure without being told what to look for. Use cases: clustering (customer segmentation), dimensionality reduction (PCA), anomaly detection. **Reinforcement learning** trains an agent through trial and error in an environment. The agent receives rewards or penalties based on actions. Use cases: game playing, robotics, recommendation optimization. ### Key ML Concepts You Must Know - **Training vs. inference**: training is the process of fitting a model to data; inference is using a trained model to make predictions - **Overfitting vs. underfitting**: overfitting memorizes training data (poor generalization); underfitting fails to capture patterns (poor accuracy) - **Cross-validation**: technique to evaluate model performance by splitting data into folds - **Feature engineering**: transforming raw data into features that improve model performance - **Confusion matrix terms**: precision, recall, F1 score, accuracy — know when each metric matters 💡 **Exam Tip:** Questions about "which type of ML" often include a clue in the scenario. If the data has labels, it is supervised. If you are finding groups or anomalies without known labels, it is unsupervised. If an agent is learning through rewards, it is reinforcement learning. --- ## Domain 2: OCI AI Services Oracle offers a suite of pre-built AI services that let developers add AI capabilities to applications without building or training custom models. Each service has a specific function — the exam tests whether you know which service solves which problem. ### OCI Vision Analyzes images and video. Capabilities include: - Object detection (identify objects and their location in an image) - Image classification (assign category labels to images) - Document AI (extract text, tables, and key-value pairs from documents) - Face detection Use when: you need to process images or documents at scale without building a custom computer vision model. ### OCI Speech Converts speech to text (transcription). Supports batch transcription of audio/video files stored in Object Storage and real-time transcription via streaming. Supports multiple languages. Use when: you need to transcribe call center audio, meeting recordings, or build voice-to-text features. ### OCI Language Natural language processing for text analysis. Capabilities: - Sentiment analysis (positive/negative/neutral/mixed at document and sentence level) - Key phrase extraction - Named entity recognition (people, places, organizations, dates) - Text classification - Language detection - Translation (via OCI Language Translation) Use when: you need to analyze customer feedback, categorize support tickets, or extract structured data from unstructured text. ### OCI Anomaly Detection Detects anomalies in time-series data using machine learning. You provide multivariate time-series training data, the service trains a model, and then you send data for inference. Uses algorithms including MSET (Multivariate State Estimation Technique) and OCSVM (One-Class SVM). Use when: monitoring IoT sensor data, detecting unusual patterns in metrics, or identifying fraud in transaction streams. ### OCI Digital Assistant A platform for building conversational AI (chatbots and voice assistants). Key concepts: - **Intents**: what the user wants to do (e.g., "book flight", "check balance") - **Entities**: specific information extracted from utterances (e.g., dates, locations, product names) - **Skills**: individual bots focused on a specific task - **Digital Assistants**: orchestrate multiple Skills Use when: building customer service bots, internal HR assistants, or voice-enabled applications. 💡 **Exam Tip:** OCI Digital Assistant is about building conversational interfaces. OCI Language is about analyzing text. They are different services. The exam may try to blur this line. --- ## Domain 3: Generative AI and Large Language Models This is a fast-moving area and one of the most important for the exam. ### What Are Large Language Models (LLMs)? LLMs are neural networks trained on massive text corpora that can generate, summarize, translate, and reason over text. They work by predicting the next token (word/subword) in a sequence. **Key LLM concepts:** - **Tokens**: the units of text LLMs process (roughly 3/4 of a word on average) - **Context window**: the maximum number of tokens the model can consider at once - **Temperature**: controls randomness in output (low = more deterministic, high = more creative) - **Prompting**: crafting input text to guide model behavior - **Prompt engineering**: systematic techniques like zero-shot, few-shot, chain-of-thought prompting ### OCI Generative AI Service Oracle's managed service for accessing LLMs on OCI. Key points: - Provides access to models from **Cohere** (Command, Embed) and **Meta** (Llama 2/3) - Offers a **Playground** for interactive experimentation - Two deployment modes: **serverless** (shared infrastructure, pay-per-token) and **dedicated AI clusters** (private GPU clusters for production or fine-tuning workloads) - Supports both **chat/generation models** and **embedding models** ### Retrieval-Augmented Generation (RAG) RAG is a pattern that combines LLMs with a knowledge base to answer questions about private or up-to-date information the model was not trained on. The flow: 1. User asks a question 2. The question is converted to an embedding (vector) 3. A vector database (e.g., OCI OpenSearch) retrieves relevant document chunks 4. Retrieved context + question are sent to the LLM 5. LLM generates a grounded answer ### Embeddings Embeddings are vector representations of text. Similar texts have vectors that are close together in vector space. Embedding models (like Cohere Embed) are used for semantic search, RAG, and clustering. ### Fine-Tuning Fine-tuning adapts a pre-trained LLM to a specific task or domain by training it further on a smaller, task-specific dataset. On OCI, fine-tuning requires a dedicated AI cluster. Use fine-tuning when: prompt engineering is insufficient and you need the model to learn domain-specific tone, terminology, or task format. --- ## Domain 4: AI Infrastructure and Data Science ### OCI Data Science Service A managed platform for the full data science lifecycle: - **Notebooks**: managed JupyterLab environments (choose compute shape, storage) - **Model catalog**: store, version, and share trained models - **Model deployment**: deploy models as REST API endpoints for real-time inference - **Jobs**: run batch ML training jobs on managed compute - **Pipelines**: orchestrate multi-step ML workflows - **AutoML**: automated feature engineering, algorithm selection, and hyperparameter tuning OCI Data Science is for custom ML workloads where you bring your own code, data, and algorithms. This is different from OCI AI Services, which are fully managed, pre-trained APIs. ### AI Infrastructure OCI provides GPU-optimized compute shapes (A10, A100 series) for training large models. These are used when: - Training custom deep learning models - Fine-tuning LLMs on dedicated clusters - Running inference at scale --- ## 4-Week Study Plan ### Week 1: AI/ML Foundations + OCI Overview - Study supervised, unsupervised, and reinforcement learning concepts - Review key ML metrics and terminology - Complete Oracle's free OCI AI Foundations learning path on Oracle University - Take notes on each OCI AI service (Vision, Speech, Language, Anomaly Detection, Digital Assistant) ### Week 2: OCI AI Services Deep Dive - For each AI service, understand: what it does, what data it takes as input, what it outputs, and when to use it - Practice identifying which service matches a given scenario - Take your first practice quiz (aim for 20 questions from Domain 2 content) ### Week 3: Generative AI and Data Science - Study LLM concepts: tokens, embeddings, fine-tuning, RAG, prompting techniques - Understand OCI Generative AI service: available models, serverless vs dedicated clusters - Study OCI Data Science: notebooks, model catalog, deployment, Jobs - Take 20-question practice quiz on Domains 3 and 4 ### Week 4: Full Practice Exams + Review - Take 2-3 full 60-question practice exams under timed conditions - Review every wrong answer — understand why the correct answer is right - Focus review on areas where you score below 70% - Re-read any topics that feel shaky 💡 **Exam Tip:** The free Oracle University learning path for this exam is excellent and aligns closely with exam content. Complete it before taking any practice exams. --- ## Top Tips for Exam Day 1. **Read every scenario carefully** — the key differentiator is usually in a single sentence (e.g., "pre-built" vs "custom model" points to AI Services vs Data Science) 2. **Know the service boundaries** — OCI Digital Assistant builds chatbots; OCI Language analyzes text; they are not interchangeable 3. **Generative AI questions are growing** — expect a third of the exam to touch LLMs, embeddings, RAG, or the OCI Generative AI service 4. **Dedicated vs serverless** is a favorite distinction — dedicated AI clusters are for fine-tuning and production isolation; serverless is for exploration and low-volume inference 5. **Do not overthink** — this is a Foundations exam; answers that are technically possible but overly complex are usually wrong --- ## Practice with 340 Questions The best way to prepare for any Oracle certification is targeted practice with exam-format questions. Our OCI AI Foundations practice exam includes 340 questions covering all four domains with detailed explanations for every answer. Ready to test your knowledge? Start with our [Oracle OCI AI Foundations (1Z0-1122-25) Practice Exam](/exams/oracle-oci-ai-foundations-1z0-1122-25-340-questions) and see where you stand.