Prepare and test your skills

Prepare and test your skills

Artificial intelligence (AI), machine learning (ML), and deep learning are nested fields with different scope. AI is the broad field of systems that perform tasks such as text generation, translation, classification, and prediction. ML is the part of AI where a model learns patterns from data and makes predictions on new data. Deep learning is a specialized form of ML that trains deep neural networks with many layers and usually needs large datasets and accelerators such as GPUs and TPUs.
AI, machine learning, and deep learning differ by scope and by how they learn from data. Traditional AI classifies or predicts from historical data, such as spam detection or house-price prediction, while generative AI creates new text, images, audio, or code. ML creates a model from examples, so data flows from datasets into a training process and then into a stored model. Supervised models learn from labeled examples, unsupervised models find structure in unlabeled data, and reinforcement models learn from feedback on actions.
Pre-trained models let teams reuse work instead of training from scratch. Model Garden (opens in a new tab) offers Google models such as Gemini and Imagen, open models such as Gemma and Llama, and third-party models from partners such as Anthropic. Common pre-trained APIs add intelligence without building models, including Cloud Vision API for images, Cloud Natural Language API for text meaning, Speech-to-Text API for audio to text, Video Intelligence API for video, Document AI API for documents, Dialogflow API for chatbots, and Gemini API for general-purpose AI. Learning transfer applies knowledge from a pre-trained model to a new task with less data and compute.
The ML lifecycle is the ordered process that turns data into business predictions. Data provides the examples, training builds the model from those examples, and inference uses the stored model to score new data. Training pays for compute to build the model and storage for the stored model, while queries against the model pay for query compute. BigQuery ML brings ML to the data in the warehouse, so analysts can create and run models with GoogleSQL queries without moving large datasets to Python frameworks.
A process flow showing data preparation and training building a stored model, then a vertical dashed inference path where new data queries the stored model to produce predictions.
The BigQuery ML workflow runs in fixed order:
Exam tip: Training builds and stores the model, while inference runs queries against the stored model to generate predictions.
The same lifecycle can run on platforms that trade control for management. Google Kubernetes Engine (GKE) (opens in a new tab) provides managed Kubernetes for large-scale training and low-latency serving with GPU and TPU flexibility. Vertex AI is a fully managed, end-to-end AI and ML platform that offloads infrastructure management, while Cloud Run is a serverless platform for containers that can scale to zero for smaller or event-driven inference.
| Service | Level of management | Choose when |
|---|---|---|
| Google Kubernetes Engine (GKE) | Managed Kubernetes with deep control | Control, custom pipelines, very large training, and low-latency serving are needed |
| Vertex AI | Fully managed end-to-end platform | Offloading infrastructure and accelerating MLOps matter most |
| Cloud Run | Serverless containers that scale to zero | Cost-effective serving of smaller models is the goal |
ML creates business value when it connects to measurable goals defined in a business use case. A value-driven approach starts from outcomes such as increased revenue, reduced costs, faster time-to-market, and higher satisfaction, then works backward to the AI solution. Reusing datasets, features, and trained models accelerates development and improves return on investment while keeping quality, security, and compliance intact.
Human-led insights and autonomous learning differ in who produces the conclusion from data. Human-led insight depends on a person who explores dashboards, profiles, and visual summaries and makes the decision. Autonomous learning depends on a trained model that learns relationships directly from data and improves its classifications, predictions, or generated content over time.
| Approach | What produces insight | Who decides |
|---|---|---|
| Human-led analysis | Queries, profiles, and visual summaries a person reads | Person interprets result |
| Autonomous learning | Classifications or generated content from a trained model | Model produces output while people govern it |
Retrospective analysis is backward-looking explanation, while predictive modeling is forward-looking forecasting. Retrospective analysis uses descriptive analytics in BigQuery to aggregate stored records into trends, such as why sales dropped last quarter. Predictive modeling with BigQuery ML trains where the data lives and then runs inference on new data to forecast demand or flag fraud. Structured data in rows and tables supports reporting with fixed rules, while unstructured data such as text, images, audio, and video requires AI that learns patterns from large training sets.
Pattern discovery and unstructured data insights are problems where machine learning succeeds when fixed rules do not. Clustering finds natural groupings in unlabeled data, such as grouping customers by shared behavior, which differs from classification that assigns data to known types. Classification assigns data to discrete categories by learning from labeled examples, such as email as spam or not spam, while regression estimates a continuous number such as price or demand.
| Task type | Output type | Example |
|---|---|---|
| Classification | Discrete category label | Email is spam or not spam |
| Regression | Continuous numerical value | Predicted house price from features and location |
Document AI is a document processing platform that transforms unstructured documents into structured data. A document file flows to a Document AI processor, the processor applies an ML model, and the service returns Document objects with extracted fields and entities. Document AI can digitize with optical character recognition (OCR), extract entities and key-value pairs such as Name: Jill Smith, and classify types. Each Google Cloud project must create its own processor instances, and results can flow to Cloud Storage, BigQuery, and search systems.
High-quality, accurate data is essential because a model learns patterns from whatever examples it receives. If training data is inaccurate, incomplete, noisy, biased, or unrepresentative, predictions will be unreliable or unfair for missing cases. Diverse data that captures many conditions helps the model generalize to new or less common examples after deployment. The Garbage In, Garbage Out (GIGO) principle states that bad inputs produce unreliable predictions because the model does not automatically correct bad data.
| Data problem | Effect on model | Prevention |
|---|---|---|
| Missing or incomplete values | Less useful training | Correct gaps and keep consistent intervals |
| Noise, errors, or inconsistency | Model learns wrong patterns | Correct errors and unify formatting |
| Data leakage | Strong test scores but poor production | Remove features unavailable at serving time |
| Training-serving skew | Poor production quality | Use only features available in same form at serving |
| Biased data | Poor generalization and unfair decisions | Capture diversity and review fairness |
A dataset split into training, validation, and test sets gives each part a different role. If splits are not specified, Agent Platform uses 80% for training, 10% for validating, and 10% for testing. The training set teaches parameters such as neural network weights, the validation set tunes hyperparameters that define structure, and the test set provides a final check on unseen data. Recommended minimums scale with complexity, including 50 rows times features for classification and 200 times features for regression.
Responsible AI is the practice of keeping automated systems fair, transparent, and trustworthy through data quality, measurement, and governance. Sources of bias include problems in training data, such as irrelevant features or poor quality, and weaknesses in design and governance, such as lack of transparency about how a model reaches its output. Reducing bias therefore requires better data feeding the model and stronger measurement and governance around the model.
Vertex Explainable AI is a feature of Vertex AI that works alongside a deployed model to interpret inferences. When the model predicts for a new input, Vertex Explainable AI also produces information that helps a person interpret the result. Feature importance shows which inputs most influenced a specific inference, while model behavior describes how the model generally responds across many inferences. Fairness metrics measure whether outcomes are fair across groups, and data governance controls such as Knowledge Catalog, BigQuery sharing, and Sensitive Data Protection improve quality and limit access to sensitive data.
Continue with this study guide: AI and ML Fundamentals
The range of Google Cloud AI and ML solutions includes pre-trained APIs, Vertex AI platform services, and custom model development tools. The three groups differ in who trains the model and how much work the user does. Pre-trained APIs need no training, Vertex AI manages the full model lifecycle, and custom tools learn from organization-specific data.
Core categories of Google Cloud AI solutions separate ready-made intelligence from models that learn from customer data. Pre-trained APIs are Google-managed services where data flows from the application to the API and back as a finished result. The application depends on the API for inference but does not manage training data or infrastructure. Vertex AI connects pre-trained capabilities and custom development in one managed environment where deployed models receive inference requests and return outputs to applications.
Common pre-trained APIs include:
Custom model development tools train on customer labels when pre-trained models are not accurate enough. Deep Learning VM Images (opens in a new tab) and Deep Learning Containers provide preconfigured environments for deeper custom work. AutoML Vision handles image classification and object detection, AutoML Natural Language handles custom classification and entity extraction, and BigQuery ML trains and runs models with GoogleSQL inside the warehouse.
Pre-built services compared with custom workflows is a choice between calling an API and building a unique model. Pre-built services accept a standard data type such as image, audio, text, video, or document and return a result without training. Data flows from the application to the service API and back, so the application depends on the service but avoids managing training. Custom workflows move data from storage into a development environment, run training on accelerated compute such as Cloud TPU, and deploy the resulting model for use.
| Option | Data and customization | Expertise and cost |
|---|---|---|
| Pre-built AI services | Standard data types and small to moderate volumes; limited customization | Application skills, not training skills; lower upfront effort and faster value |
| Custom ML workflows | Large volumes of relevant training data; high customization | Machine learning expertise; higher effort for setup, training, and tuning |
Custom environments include TensorFlow Enterprise for building TensorFlow models, Colab Enterprise for collaborative notebooks, Deep Learning VM for preconfigured virtual machines, and Deep Learning Containers for consistent containers. Choose pre-built services when expertise is limited and speed matters, and choose custom workflows when no API matches the problem or large volumes of specialized data justify extra effort.
Mapping business use cases to products means matching data type, customization need, and desired outcome. Document AI handles end-to-end document understanding, specialized APIs handle single tasks like vision or language, Vertex AI unifies custom training and inference, and BigQuery ML brings prediction directly to warehouse data. Vertex AI capabilities include batch inference that processes a group of requests into one output file, online inference that returns real-time results, Imagen for text-to-image generation and editing, and Vertex AI Search for finding relevant documents in a corpus.
Document processing converts dark data trapped in emails, PDFs, scanned forms, and receipts into usable fields. The pipeline runs in fixed order:
Optical character recognition must run first and classification runs second on the resulting plain text. The Cloud Natural Language API classifies with general categories and needs no training, while AutoML Natural Language trains on business labels.
Exam tip: Speech-to-Text converts audio to text, while Text-to-Speech converts text to audio — the direction of the flow determines the service.
Decisions and tradeoffs for Google Cloud AI and ML selection balance speed, effort, differentiation, and required expertise. Fully managed services and pre-trained models shorten deployment timelines, while self-managed infrastructure and custom training give deeper control but take longer. Organizations choose well by matching the level of management to how fast they need results and what skills they can operate over time.
| Service | Level of management | Best for |
|---|---|---|
| Vertex AI | Fully managed, end-to-end | Teams focused on MLOps and rapid time-to-value |
| Cloud Run | Serverless, scales to zero | Event-driven applications and smaller models |
| GKE | Self-managed Kubernetes | Deep control, portability, and custom pipelines |
Vertex AI (opens in a new tab) offloads infrastructure so teams focus on models and reach value faster. Model Garden offers pre-trained models grouped as Google models such as Gemini and Imagen, open models such as Gemma and Llama, and third-party models from partners such as Anthropic. Fine-tuning sits in the middle because learning transfer applies knowledge from a pre-trained model to a new task with less data and compute. GKE is chosen when the organization needs infrastructure control, large-scale training with minimal-latency serving, or portability with open standards, but that control requires more setup and Kubernetes expertise.
Continue with this study guide: Google Cloud’s AI and ML solutions
Choosing between pre-trained APIs, AutoML, and custom training is how organizations get business value from Google Cloud AI at the right cost, speed, and accuracy. Pre-trained APIs are fastest when a general model is good enough, AutoML builds a tailored model from company data without deep coding, and custom training gives full control for specialized needs.
Strategic use of pre-trained APIs means calling ready-to-use models instead of building models from scratch. Data flows from the application to the API, and the AI result flows back from the API to the application, while Google manages the model, training data, and infrastructure. Model Garden supports model selection with Google models like the Gemini family and Imagen, open models like Gemma and Llama, and third-party models from partners like Anthropic.
Common Google Cloud APIs cover standard jobs without custom training:
This approach reduces time-to-market from months to days and reduces the need for data science experts. A single application can combine several APIs, such as sending a document to one API for extraction and then sending the extracted text to another API for analysis.
The choice depends on whether an existing model is good enough, whether training data is available, and how much control the team needs. First check if a pre-trained API meets the requirement; if it does, use it to save time and cost. AutoML in Vertex AI is a managed training option that builds a custom model from an organization's own labeled data without extensive coding, so it captures patterns that a general model trained on public data misses.
| Option | Data needed | Expertise and control |
|---|---|---|
| Pre-trained API | No training data needed | Lowest effort, no control over model internals, best when general accuracy is enough |
| AutoML | Requires sufficient labeled proprietary data | Low expertise needed, managed tuning, best for higher accuracy without custom code |
| Custom training | Requires training data plus ML engineering capacity | Highest control over algorithm, layers, learning rate, hyperparameters, and explainability |
Custom training fits when the team needs detailed control and explainability, while a specialized product like Anti Money Laundering AI shows differentiation because it trains only on the customer's own banking data to generate monthly risk scores with explainability for investigation.
BigQuery ML is a capability of BigQuery that lets analysts build and run machine learning models directly inside the data warehouse using standard SQL. A model created with BigQuery ML is stored in a BigQuery dataset like a table or view, so training data never has to leave the warehouse for a separate Python environment. Data locality means bringing machine learning to the data instead of moving data to the model, which reduces complexity and avoids data transfer costs and latency.
The SQL workflow in BigQuery ML is the ordered path from stored tables through training to prediction without leaving BigQuery:
CREATE MODEL statement to train a model on a query result.TRANSFORM functions and hyperparameter tuning when needed.The three core statements are CREATE MODEL to train, ML.EVALUATE to check quality, and ML.PREDICT to generate predictions. Regression and classification models predict a label from labeled examples, K-means clustering segments data without labels, matrix factorization builds recommendations, and time series models like ARIMA_PLUS forecast and detect anomalies.
| Function | Requires stored model | Input |
|---|---|---|
| ML.EVALUATE | Yes | No input returns training-time metrics; specified input tests on new data |
| ML.METRICS | No | Any table or query with actual and predicted columns |
Selecting the correct Google Cloud pre-trained API means matching the job — audio direction, image task, or text task — to the service built for that job. All of these APIs are pre-trained, so developers add the capability with an API call without training a custom model. Audio APIs are chosen by conversion direction, image APIs by what must be found in the picture, and text APIs by whether the task is analysis or translation.
| API | Conversion direction | Choose when |
|---|---|---|
| Speech-to-Text | Audio to text | Input is spoken audio and output must be written text, such as voicemail to message or captions |
| Text-to-Speech | Text to audio | Input is written text and output must be spoken audio, such as phone assistants and accessibility readers |
The Cloud Vision API is the pre-trained API that analyzes images without custom training. An application sends an image from a local file or Cloud Storage with a requested feature type, and the API returns annotations describing what it found. LABEL_DETECTION returns objects, scenes, or activities, OBJECT_LOCALIZATION detects multiple objects and where each sits, FACE_DETECTION finds faces, LOGO_DETECTION finds logos, and SAFE_SEARCH_DETECTION screens for unsafe content. For text in images, TEXT_DETECTION is optimized for sparse text such as a sign, while DOCUMENT_TEXT_DETECTION is for dense documents, PDF or TIFF files, and handwriting and takes precedence when both are requested.
The Cloud Natural Language API analyzes text structure and meaning, while the Cloud Translation API translates text between languages. Cloud Translation - Basic API offers only global endpoints, AutoML Translation offers only global endpoints, and Cloud Translation - Advanced API offers both global endpoints and regionalized endpoints configurable for data location. A business that must keep translation data inside a region therefore needs Cloud Translation - Advanced API with a regionalized endpoint.
Exam tip: Speech-to-Text converts audio to text, while Text-to-Speech converts text to audio — check the starting input to pick the correct API.
Custom models on Vertex AI are machine learning models trained or adapted on an organization's own data to solve its specific problems. A generic pre-trained solution handles common tasks the same way for everyone, while a custom model learns unique patterns, language, and customer behavior that competitors using only generic models cannot copy. Proprietary data such as purchases, product images, sales history, and support tickets teaches the model the patterns of that specific business.
The Vertex AI Model Development Lifecycle is the ordered workflow from preparing data to serving predictions:
Deployment in Vertex AI is placing a trained model into production for inference on unseen data. Vertex AI Model Registry is the central repository where teams organize versions and deploy a chosen version directly or by alias to an endpoint. Batch inference processes a group of requests together into one output file, while online inference returns real-time results for individual data points as they arrive.
| Option | Data used | Strategic effect |
|---|---|---|
| Generic pre-trained API | Broad public data | Fast to launch but the same for every competitor |
| Custom model on Vertex AI | Organization's own image, text, tabular, and video data | Harder to build but creates a capability competitors do not have |
| Fine-tuned foundation model | Broad base plus company examples | Balances strong base quality with business-specific voice and knowledge |
Cloud Run (opens in a new tab) hosts AI applications by deploying code or a container as a scalable API endpoint where the application sends data to a model and returns results. Value improves through an iterative cycle of monitoring indicators and costs, making data-driven adjustments, refining objectives, and reviewing results with stakeholders. MLOps scales this reliably with version control, continuous integration and delivery with Cloud Build (opens in a new tab) and Artifact Registry, observability with Cloud Monitoring and Cloud Logging, and autoscaled serving.
TensorFlow is end-to-end open source software for building and training machine learning models, while Cloud TPU is Google's proprietary hardware that accelerates those models. TensorFlow provides the tools from preparation to deployment across many different systems, and Cloud TPUs provide the specialized math performance that runs TensorFlow jobs faster than general-purpose processors. Google open sourced TensorFlow in 2015, and TensorFlow Enterprise is Google Cloud's enterprise version with support and faster access from cloud storage to training.
Tensor Processing Units (TPUs) are Google's custom application-specific integrated circuits designed to accelerate ML and AI workloads. A TensorFlow job sends heavy math operations to the TPU, and the TPU returns results much faster than general-purpose compute. Three hardware capabilities drive this acceleration:
Provisioning TPUs controls availability and scaling through managed services. Compute Engine creates and manages individual TPU VMs or slices with full lifecycle management, while Google Kubernetes Engine (GKE) orchestrates TPU nodes and node pools inside Kubernetes clusters for large-scale AI workloads.
| Option | What it manages | Choose when |
|---|---|---|
| Compute Engine | Individual TPU VMs or slices | The team needs direct control over TPU VM lifecycle |
| Google Kubernetes Engine (GKE) | TPU nodes and node pools | The workload needs Kubernetes orchestration across many AI jobs |
Continue with this study guide: Building and using Google Cloud AI and ML solutions
CREATE MODEL to train, ML.EVALUATE to check quality, and ML.PREDICT to generate predictions without moving data.TEXT_DETECTION for sparse text and DOCUMENT_TEXT_DETECTION for dense documents and handwriting.Artificial intelligence is the broad field of systems that perform tasks such as text generation, translation, classification, and prediction. Machine learning is the part of artificial intelligence where a model learns patterns from data and makes predictions on new data, while deep learning is a specialized form of machine learning that trains deep neural networks with many layers and usually needs large datasets and accelerators such as GPUs and TPUs.
Pre-built AI services are best for standard data types such as image, audio, text, video, or document in small to moderate volumes with limited customization. They require application skills rather than training skills for lower upfront effort and faster value, while custom ML workflows are best when large volumes of relevant training data and high customization justify higher effort for setup, training, and tuning with machine learning expertise.
Training builds the model from examples and stores it, paying for compute to build the model and storage for the stored model, while inference uses the stored model to score new data and pays for query compute. In BigQuery ML, training uses CREATE MODEL to train without moving data and inference uses ML.PREDICT to generate predictions on new data.
Last updated