Estonian defence company KrattWorks builds aerial target drones for the Air Defence Battalion. The aerial target is powered by a petrol engine and has a wingspan of 2.2�m without the body: only the wing reaches a maximum speed of 180�km per hour and remains airborne for around 20 minutes. In addition to smoke torches, the drones are equipped with heat elements manufactured by Ruf Eesti that can emit heat up to 2,000 degrees The drone is designed to serve as a target for machine guns, anti-aircraft guns and short-range anti-aircraft missiles.

In order to improve the searchability of the necessary information from the Riigteataja (riigiteataja.ee) and to simplify the understanding of the information, a pilot project was carried out to train large language models on the data of the Riigteataja containing national legislation and thereby create a customized model. As part of the pilot project, testing was done in the Azure Open AI studio. The language model tested was the large language model of ChatGPT 3.5 Turbo.

The purpose of the project was to test whether machine learning can be used to determine the risk level of unemployment for individuals, as well as to predict this risk, provide an explanation for the risk level and develop a decision support application to improve the targeting of prevention services.

Fire prevention: risk level forecasting.

The aim of the healthcare application is to develop algorithms for translating Estonian health data into the internationally recognised standard (OMOP-CMD) and to test algorithms on the obtained data to determine the risks of different diseases. In the future, this will facilitate the use of algorithms developed elsewhere to more accurately determine the health risks of the residents of Estonia.

The aim of the cyber security application was to create algorithms for automated incident detection, filtering and prioritisation that would allow specialists to quickly identify the threats critical to their organisation, without having to spend excessive amounts of time monitoring large streams of data.

Machine translation was implemented in the Ametlikud Teadaanded, which translates the information on the website into English. The main objective of implementing machine translation is to support Estonia�s open business environment and reduce translation costs. Among other benefits, the solution ensures that the information is understandable to foreigners on company boards or foreign creditors. Machine translation uses a neural network learning algorithm, allowing it to improve over time. The publisher monitors the quality of the translated texts and makes corrections as necessary. Machine translation learns from these corrections and thus the quality of translations is improved in the future. Only the Estonian text of a notice is considered authentic and as having legal meaning.

Within the project, a proof of concept solution was developed for a kratt supporting individual pathways of students. During the project, the application of machine learning was tested in the development of recommendation models for educational materials. In addition, prototype-like models were created to help automate the personalisation of learning materials and activities based on the learners' interests, knowledge and abilities.

As part of the project, the principles and architectural vision of the basic and alternative processes of an automated article keywording solution were developed along with the prerequisites and requirements necessary for the implementation of the solution, a description of the future situation, a cost-benefit analysis and risk assessment.

The purpose of the project was to ensure that the most affordable medicinal products are sold in pharmacies, so that patients always have the opportunity to choose the most affordable solution. During the project, an analysis was conducted to determine which data could be used to develop a risk model for assessing the availability of price agreements on medicinal products. Subsequently, a corresponding software prototype was created. The machine learning-based risk model assists inspectors at the Agency of Medicines in the preparation of a risk-based inspection plan.

This is a prototype of Estonian speech synthesis based on neural networks that has been developed by the Natural Language Processing research group at the University of Tartu, and is trained on the corpus of Estonian news. Speech synthesis can currently imitate the voices of six different speakers. The project is still in the development stage and is far from perfect; however, the neural network-based speech synthesis sounds more natural than earlier methods. The strengths of the speech model include the natural sound and intonation of speech, and the pronunciation of numbers, symbols and abbreviations.

This is a prototype of Estonian speech synthesis based on neural networks that has been developed by the Natural Language Processing research group at the University of Tartu, and is trained on the corpus of Estonian news. Speech synthesis can currently imitate the voices of six different speakers. The project is still in the development stage and is far from perfect; however, the neural network-based speech synthesis sounds more natural than earlier methods. The strengths of the speech model include the natural sound and intonation of speech, and the pronunciation of numbers, symbols and abbreviations.