AI chatbots for researchers within the CESNET environment
AI chatbots operated within the CESNET / e-INFRA CZ infrastructure allow users to work with various large language models within the Czech academic infrastructure.
Unlike publicly available AI services, these services operate in a closed environment, providing a higher level of control over the data being processed. In addition, the data are not shared with third parties. They can therefore also represent a suitable alternative for users who would otherwise not have access to paid AI services.
How can chatbots help researchers?
Chatbots can be useful for common research and academic tasks, for example:
- summarizing academic texts,
- working with programming code,
- extracting text from documents, or
- preparing and anonymizing data.
Examples of use based on testing by the MENDELU Open Science Centre
During preliminary testing by the MENDELU Open Science Centre, different models proved suitable for different types of tasks. For example, the Qwen 3.5 and Mistral Medium 3.5 models were tested for transcribing and processing handwritten text. The Thinker model was used for thematic coding of interviews and for simplifying academic articles. The Kimi K3 model was tested for analytical tasks, while the Coder model was used for working with programming code. Image generation capabilities were also tested using the GPT OSS 120B model.
These examples are based solely on practical testing by the MENDELU Open Science Centre and should not be understood as universal recommendations, as both the models and their availability continue to evolve.
The individual available models differ in their capabilities, and it is therefore advisable to select a model according to the specific type of task. Before using them, we recommend reviewing the documentation describing the features and limitations of the individual models. Although the environment is operated locally, it is still advisable to carefully consider what data are entered into the system, particularly in the case of unpublished results, personal data, or other sensitive information.
The detailed guide [in Czech] prepared by the MENDELU Open Science Centre describes registration and signing in to the service, selecting an AI model, as well as other available features, including working with an API key, web search, and image generation.
In addition to AI chatbots, the e-INFRA CZ infrastructure also offers other tools that can be used in research, such as Jupyter Notebooks, Claude Code, and DeepSite for generating websites from text prompts.