O R A C L E C E R T I F I C A T I O N
1Z0-1122 Oracle Cloud Infrastructure AI Foundations 2022 Associate Practice Exam
Exam Number: 4825 | Last updated April 19, 2026 | 700+ questions across 4 vendor-aligned objectives
The 1Z0-1122 Oracle Cloud Infrastructure AI Foundations 2022 Associate exam is an entry-level credential for technical and non-technical learners who need a working mental model of artificial intelligence on OCI. Candidates confirm conversational fluency in machine-learning fundamentals, deep learning and generative AI concepts, and the OCI AI services that make those models accessible to applications.
The heaviest content is AI and Machine Learning Fundamentals (roughly 35%), covering supervised, unsupervised, and reinforcement learning, evaluation metrics, bias and variance, and the project lifecycle from problem framing to deployment. Deep Learning and Generative AI contributes another 25% with neural network basics, large language models, and embeddings.
OCI AI Services sits near 25% and drills into OCI Language, OCI Vision, OCI Speech, OCI Document Understanding, and the prebuilt generative AI service. OCI Data Science and Responsible AI rounds out the remaining weight with the notebook and model-deployment model and the ethics and governance principles every AI project should follow.
Every answer links to the source. Each explanation below includes a hyperlink to the exact Oracle documentation page the question was derived from. PowerKram is the only practice platform with source-verified explanations. Learn about our methodology →
946
practice exam users
95.0%
satisfied users
94.2%
passed the exam
4.5/5
quality rating
Test your 1Z0 1122 OCI AI Foundations Associate knowledge
10 of 700+ questions
Question #1 - AI and Machine Learning Fundamentals
An insurance analyst wants to predict whether a new claim is fraudulent based on 20 historical features including claim amount, claimant history, and provider code. She has 50,000 labeled claims (fraud or not fraud).
What type of machine learning task is this?
A) Supervised learning — binary classification.
B) Reinforcement learning.
C) Regression.
D) Unsupervised clustering.
Show solution
Correct answers: A – Explanation:
With labeled examples and a two-class output (fraud vs not fraud), this is supervised binary classification. Option D has no labels. Option B involves sequential decisions with rewards. Option C predicts a continuous value, not a category. Source: Check Source
Question #2 - AI and Machine Learning Fundamentals
A data scientist has built a classifier with 99% accuracy on the training set but only 72% on the test set. On new production data, accuracy drops further.
Which concept best describes this pattern?
A) Bias in the data labels.
B) Overfitting — the model memorized training data and does not generalize.
C) Underfitting — the model is too simple.
D) Bitrot in the model file.
Show solution
Correct answers: B – Explanation:
High train accuracy with much lower test and production accuracy is the hallmark of overfitting. Option C would show low train accuracy too. Option A describes label quality issues, not the gap between train and test. Option D is not a recognized ML concept. Source: Check Source
Question #3 - Deep Learning and Generative AI
A product manager describes a chatbot that answers customer questions using recent support documentation. Which technology family best fits?
Which AI approach matches?
A) K-means clustering of customers.
B) A linear regression on ticket volume.
C) Large language models combined with retrieval-augmented generation.
D) A simple decision tree with no text input.
Show solution
Correct answers: C – Explanation:
LLMs handle natural language, and RAG grounds them in current documentation by retrieving relevant passages before generating answers — the canonical match. Option D cannot generate language. Option A and Option B are not language models. Source: Check Source
Question #4 - OCI AI Services
A business analyst wants to extract sentiment and named entities from thousands of customer feedback comments without training a custom model.
Which prebuilt OCI AI service fits?
A) OCI Vision for images.
B) OCI Speech for audio.
C) OCI Document Understanding for scanned forms.
D) OCI Language for sentiment analysis and entity recognition.
Show solution
Correct answers: D – Explanation:
OCI Language is purpose-built for NLP tasks including sentiment and named-entity recognition on text, with prebuilt models. Option A is for images. Option B is for audio. Option C is for structured document extraction. Source: Check Source
Question #5 - OCI AI Services
A logistics provider wants to extract key-value pairs like invoice number, date, and total from scanned vendor invoices, without building a custom OCR pipeline.
Which OCI AI service fits key-value extraction from scanned vendor invoices?
A) OCI Document Understanding with key-value extraction.
B) OCI Language sentiment.
C) OCI Speech text-to-speech.
D) OCI Vision object detection only.
Show solution
Correct answers: A – Explanation:
OCI Document Understanding is designed for structured document extraction including key-value pairs from invoices and forms. Options B, C, and D target different modalities or task types. Source: Check Source
Question #6 - Deep Learning and Generative AI
A developer is building a semantic search that finds similar documents to a query. She converts each document and query to a fixed-length numeric vector and searches by similarity.
What is the name of these numeric vectors?
A) Decision boundaries.
B) Embeddings.
C) Hashes.
D) Learning rates.
Show solution
Correct answers: B – Explanation:
Embeddings are dense vectors that capture semantic meaning, enabling similarity search. Option C is a deterministic digest that does not preserve semantic similarity. Option A is a classifier concept. Option D is an optimizer hyperparameter. Source: Check Source
Question #7 - AI and Machine Learning Fundamentals
A team is building a churn model. The classes are imbalanced: 95% retain, 5% churn. They report 95% accuracy but miss almost every churner.
Which evaluation metric would better reflect the churn detection quality?
A) R-squared.
B) Mean squared error.
C) Recall on the churn class (or F1 for the minority class).
D) Overall accuracy only.
Show solution
Correct answers: C – Explanation:
With class imbalance, overall accuracy hides minority-class misses; recall (or F1) on the minority class captures churn-detection quality. Option D is what misled the team. Option B and Option A are regression metrics, not classification. Source: Check Source
Question #8 - Data Science and Responsible AI
An AI ethics lead insists that every model decision affecting a customer should be explainable, with documented training data lineage and a process to investigate complaints.
Which responsible AI principles does this describe?
A) GPU utilization.
B) Speed and throughput.
C) Model size and parameter count.
D) Transparency and accountability.
Show solution
Correct answers: D – Explanation:
Transparency (explainability, lineage) and accountability (complaint process) are core responsible-AI principles. Options A, B, and C are performance or engineering metrics, not ethics principles. Source: Check Source
Question #9 - OCI AI Services
A call center wants real-time transcripts of customer calls so agents can highlight keywords and supervisors can audit handling.
Which OCI AI service fits real-time audio transcription for call center handling?
A) OCI Speech for audio-to-text transcription.
B) OCI Vision object detection.
C) OCI Document Understanding key-value extraction.
D) OCI Language sentiment only.
Show solution
Correct answers: A – Explanation:
OCI Speech provides audio-to-text transcription, the necessary foundation for real-time call transcripts. Option B is images. Option D processes text, not audio. Option C is for documents. Source: Check Source
Question #10 - Deep Learning and Generative AI
An engineer wants an LLM response that is both creative and diverse for brainstorming. She notices responses are too repetitive and deterministic with default settings.
Which generation parameter should she increase to encourage more diverse outputs?
A) The number of GPUs.
B) The batch size.
C) Temperature (and/or top-p).
D) The parameter count of the model.
Show solution
Correct answers: C – Explanation:
Temperature (and top-p / nucleus sampling) control output diversity: higher temperature produces more varied, creative output. Option D changes the model, not generation behavior. Options A and B affect throughput, not diversity. Source: Check Source
Get 700+ more questions with source-linked explanations
Every answer traces to the exact Oracle documentation page — so you learn from the source, not just memorize answers.
Exam mode & learn mode · Score by objective · Updated April 19, 2026
Learn more...
What the 1Z0 1122 OCI AI Foundations Associate exam measures
- AI and machine learning fundamentals (35%) — distinguish supervised, unsupervised, and reinforcement learning, evaluate models with appropriate metrics, and frame AI projects end to end.
- Deep learning and generative AI (25%) — explain neural networks, large language models, embeddings, and the strengths and limits of modern generative systems.
- OCI AI services (25%) — select OCI Language, Vision, Speech, Document Understanding, or the prebuilt Generative AI service for a given business use case.
- Data Science and responsible AI (15%) — describe the OCI Data Science notebook and model-deployment workflow and apply fairness, transparency, and accountability principles.
How to prepare for this exam
- Review the official 1Z0-1122 exam page and capture the objectives and weights.
- Complete the Oracle University OCI AI Foundations Associate learning path on MyLearn, which is free and covers every objective.
- In an OCI tenancy, run a sentiment analysis against OCI Language, score images with OCI Vision, and try a prompt against the Generative AI service playground.
- Apply the knowledge on real conversations: explain the difference between supervised and generative AI to a stakeholder or draft a one-page use case for an OCI AI service.
- Master one objective at a time, starting with AI and machine learning fundamentals since it carries the most weight.
- Run PowerKram learn mode to see feedback after every question with sourced links back to Oracle documentation.
- Finish with PowerKram exam mode across all objectives until you pass three full-length attempts in a row.
Career paths and salary outlook
AI Foundations opens the door to AI-adjacent careers across industries:
- AI Program Analyst — $80,000–$115,000 (Glassdoor).
- Junior Machine Learning Engineer — $95,000–$130,000 (PayScale).
- AI Solution Consultant — $110,000–$150,000 (U.S. Bureau of Labor Statistics).
Official resources
Start with the OCI AI Foundations Associate Learning Path on Oracle MyLearn. Supplement with the OCI AI Services documentation and the Oracle Artificial Intelligence overview.