https:\/\/taltech.atlassian.net\/servicedesk\/customer\/portal\/14\/group\/33\/create\/374<\/a><\/p>\n3. Ordering and using services<\/strong>
After access has been created, it is possible to order and use AI\/ML services and virtual resources (computing power, data storage, etc.) provided by Azure. Pricing information is available directly in the Azure AI catalog for each model.<\/p>\n\n
What is Azure AI?<\/h3>\n
Microsoft Azure AI enables the use and development of different artificial intelligence solutions \u2014 from large language models (LLMs) to image, speech, and data analysis. All models and services are available at: https:\/\/ai.azure.com\/catalog<\/p>\n
Azure AI, also known as Microsoft Foundry, is Microsoft Azure\u2019s artificial intelligence platform that enables users to develop, test, and use different AI solutions. It provides access to services for large language models, agents, image processing, speech, translation, document processing, data analysis, and machine learning.<\/p>\n
Azure AI is mainly suitable for development, research, and data processing projects where there is a need to build or integrate AI-based solutions, use APIs, manage cloud resources, process data, or test different models in a controlled Azure environment.<\/p>\n
Models and services are available in the Azure AI \/ Microsoft Foundry model catalogue: https:\/\/ai.azure.com\/catalog<\/p>\n
Why use Azure AI?<\/h3>\n
Azure AI enables users to:<\/p>\n
\n- use state-of-the-art AI models without building dedicated GPU infrastructure;<\/li>\n
- develop AI-based applications, agents, and workflows;<\/li>\n
- use large language models for text generation, summarisation, question answering, classification, data structuring, and coding-related tasks;<\/li>\n
- process documents, images, speech, and other data;<\/li>\n
- support research, data analysis, and prototyping;<\/li>\n
- scale solutions according to project needs;<\/li>\n
- manage access, resources, costs, and security settings in a centrally managed Azure environment.<\/li>\n<\/ul>\n
What models can be used?<\/h3>\n
The Azure AI \/ Microsoft Foundry model catalogue provides access to hundreds of models from different model providers. The exact availability of models depends on the Azure region, subscription, licence, selected deployment type, and Microsoft\u2019s current service offering.<\/p>\n
Examples of model providers and model families that may be available through Azure AI \/ Microsoft Foundry include:<\/p>\n
* OpenAI \/ Azure OpenAI models, such as GPT-series models for text, code, tool use, and multimodal tasks;
\n* Anthropic Claude models, where available in the relevant region, subscription, and deployment type;
\n* Microsoft models, such as Phi and other Microsoft-developed models;
\n* Meta Llama models for open model use cases;
\n* Mistral models for text-based and multilingual tasks;
\n* Cohere models, including models for text understanding, search, and reranking use cases;
\n* DeepSeek models, including reasoning and code-related use cases;
\n* xAI Grok models, where available in the relevant region and subscription;
\n* Black Forest Labs FLUX models for image-related use cases;
\n* NVIDIA and Hugging Face models, depending on the models and deployment options available in the catalogue.<\/p>\n
Not all models are available in every region or for every Azure subscription. Before adoption, the specific model card in the Azure AI \/ Microsoft Foundry catalogue must be reviewed. The model card provides information about the model description, terms of use, pricing, regions, deployment options, and technical limitations.<\/p>\n
Why use Azure AI at the university?<\/h3>\n
In the university context, Azure AI is particularly suitable in cases where a standard chat assistant is not sufficient and there is a need to develop, integrate, or manage an AI solution at a more technical level.<\/p>\n
Main use cases include:<\/p>\n
\n- research and development projects where different AI models need to be tested;<\/li>\n
- data analysis, classification, enrichment, or structuring;<\/li>\n
- document processing, search, summarisation, and knowledge management;<\/li>\n
- creation of AI-based prototypes, applications, or services;<\/li>\n
- development of agents and automated workflows;<\/li>\n
- testing speech, translation, image, or multimodal solutions;<\/li>\n
- solutions requiring more precise control over access, resource usage, logging, monitoring, and costs.<\/li>\n<\/ul>\n
Data protection and security<\/h3>\n
When using Azure AI, Microsoft Azure\u2019s enterprise-level security, administrative, and compliance measures apply. Azure enables the use of services in selected Azure regions, including European Union regions, and supports the management of access, network connections, logging, permissions, and resource-based billing through Azure\u2019s standard security mechanisms.<\/p>\n
For Azure OpenAI and models provided by Azure through Microsoft Foundry, customer inputs and outputs are not used to train foundation models. Data is processed for service delivery, security, and abuse monitoring in accordance with Microsoft\u2019s service terms and the selected service configuration.<\/p>\n
The actual location of data processing and storage depends on the selected service, region, model, and deployment type. If a project requires data processing and storage to take place within the European Union or the European data boundary, an appropriate Azure region and deployment type must be selected, and global deployment options must be avoided where they do not meet the project\u2019s data protection or contractual requirements.<\/p>\n
Third-party and community models may be subject to separate terms of use, licences, pricing, and data processing conditions. Before using such models, the terms of the specific model must be checked in the Azure AI \/ Microsoft Foundry catalogue.<\/p>\n
Using Azure AI does not mean that all data may be entered into the service without restrictions. Only data whose processing is permitted, necessary, and appropriate for the specific work task or project may be used. If the use case involves personal data, student-related data, confidential information, contractually restricted information, or security-critical information, it must be assessed before use whether Azure AI is permitted and compliant with TalTech\u2019s requirements.<\/p>\n
Passwords, secret keys, API keys, access tokens, authentication credentials, certificate private keys, or other access credentials must not be entered into Azure AI services, except where a secure and purpose-built Azure key or secrets management service is used, such as Azure Key Vault.<\/p>\n
How to use Azure AI services<\/h3>\n
The use of Azure AI services requires an Azure subscription, access rights, and, where necessary, a corresponding cloud account or administrator role.<\/p>\n
Azure subscription opening<\/h4>\n