You Know Copilot. But What Should You Use When You Need AI for Research?

Copilot Is Not the Only AI Infrastructure Available to the MUNI Academic Community

CERIT-SC at the Institute of Computer Science of Masaryk University, part of the national e-INFRA CZ infrastructure, also offers AIaaS (AI as a Service). The service provides access to large language models operated on infrastructure located in the Czech Republic, giving researchers another option for using generative AI when working with data, text or code.

Two AI Tools, Two Different Worlds

If you need to write an email, summarise a Word document, prepare a presentation or work with content in Microsoft 365, Copilot is a natural choice. Its main advantage lies in its integration with the tools already used across the university.

For research, programming or working with your own data, however, you often need something different: the ability to control the model programmatically, connect it to your own tools, experiment with different models or work with data in an environment over which you have greater control.

This is precisely what the CERIT-SC AI infrastructure, AIaaS, is designed for.

It’s Not About Competition. It’s About Choosing the Right Tool.

Copilot has one major advantage: it is already where you work. If you have Word or Outlook open and need to quickly create, edit or summarise something, there is no additional setup required.

CERIT-SC, on the other hand, comes into its own when you do not want to simply use AI, but integrate it into your own workflows.

1) When Data Are Sensitive, Where the AI Runs Matters

One of the most important differences lies in the underlying infrastructure. The university version of Copilot provides Enterprise Data Protection. Among other things, this means that organisational data are protected from being used to train public models. For everyday work, this is an important and useful safeguard.

However, the data are still processed within Microsoft’s cloud infrastructure. CERIT-SC offers a different model: AI services run on e-INFRA CZ’s own physical infrastructure in the Czech Republic. The service is based on NVIDIA accelerator clusters and is designed, among other things, for working with research data and developing custom solutions.

This can be particularly important when working with:

• unpublished research results,
• data generated within research projects,
• internal experimental data,
• data subject to restrictions arising from grant conditions.

If it matters to you not only that your data are not used to train a model, but also where they are physically processed, CERIT-SC offers a fundamentally different option.

2) When You Want to Integrate AI into Your Own Program, Copilot Is Not Enough

Imagine you have a Python script that processes thousands of records every day. You want AI to classify the results, generate a structured output and then pass it on to the next step in the analysis. In a standard Copilot chat, you can of course ask it to write or modify the code you need. But what if you want to call the AI directly from your own program?

CERIT-SC provides an OpenAI-compatible API at llm.ai.e-infra.cz. For developers, this makes a crucial difference: the LLM does not have to remain confined to a web chat window. You can integrate it directly into your own program, script, research pipeline or application.

The same principle can be used, for example, in Python or R, in automated data processing, when developing custom applications, or when experimenting with AI agents.And this is one of the key differences: Copilot helps you work. CERIT-SC lets you integrate AI into the way you work.

3) AI Directly in Jupyter: From Question to Result Without Switching Windows

For many researchers and students, the main working environment is not Word or Excel, but JupyterLab. That is why CERIT-SC offers AI integration directly in Jupyter Notebook. AI can assist with writing, explaining, debugging and optimising code in Python or R.

Instead of the workflow: Jupyter → copy code → AI chat → copy response → back to Jupyter

you can work with AI directly in the environment where you perform your analysis.

This is useful not only for experienced programmers. For students, AI can also serve as a guide when learning to code, explaining errors or suggesting the next step directly while they work in the notebook.

4) More Than Just Chat: arXiv, GitHub, Your Own Documents and Python

CERIT-SC AIaaS is more than just another website where you type in a question. The Chat AI environment, built on Open WebUI, includes tools tailored to the needs of researchers and developers. These include working with scientific resources from arXiv, searching GitHub, running Python in a sandbox environment, and working with your own documents using RAG.

It is precisely this combination of a language model and additional tools that matters. A research question often does not simply sound like: “Write me a text about quantum physics.” It may instead be: “Find relevant papers, compare their approaches, extract the methods and results they used, and prepare the material for further analysis.” In such cases, a simple chatbot becomes a working tool for research.

5) What About New AI Models?

Another key difference lies in access to models. Copilot is built around the ecosystem of commercial Microsoft and OpenAI models. Users do not need to decide which model to choose, which can be an advantage for everyday use.

CERIT-SC takes a different approach. It provides access to powerful open-source models and allows users to work with different models depending on the specific task. The available portfolio includes, for example, recent models from the Kimi, GLM, Qwen, DeepSeek and GPT-OSS families.

This is particularly important for research and development: you can experiment, compare models and identify the solution best suited to a specific task, rather than being limited to a single predefined AI product.

6) AI for Programmers: From Code Generation to Agents

AI is no longer just about generating a few lines of Python. Modern development is increasingly moving towards so-called agentic systems – tools that can work with repositories, run commands, modify files, or interact with other services.

CERIT-SC therefore also supports tools and integrations designed for this way of working, including environments such as Claude Code, OpenCode, other agentic and IDE tools, as well as support for MCP servers.

For developers, AI no longer has to be “a window next to the editor”. It can become part of the development environment itself. This opens up opportunities for faster prototyping, code refactoring, automation of routine tasks, and the development of custom research applications.

So: Copilot or an LLM from CERIT-SC?

The simplest answer is: it depends on what you are doing.