https://www.epo.org/en/news-events/news/codefest-spring-2025-finalists

CodeFest Spring 2025 finalists

Image
CodeFest 2025

The European Patent Office (EPO) has announced the six finalists of its CodeFest Spring 2025 competition to create an automated system for classifying patent data to accelerate sustainability-focused innovation and contribute to the United Nations Sustainable Development Goals (UN SDGs)

The winners will be revealed at the prize ceremony on July 3 at 14.40 CEST during the PATLIB conference, where they will also present their proposed solutions and offer insight into the future of patent classification and sustainability.

Join us to celebrate the finalists and find out which three teams will take home a CodeFest trophy and cash prize. The grand prize winner will receive EUR 20 000, with the first and second runner-up each receiving EUR 10 000 and EUR 5 000 respectively.

Meet the finalists

This third edition of CodeFest received a total of 33 innovative proposals submitted by individuals and teams from 30 countries. The six finalists were selected through a careful evaluation conducted by a jury of EPO experts in IT, data science, AI and patent information.  The finalists are presented in alphabetical order.  

ConfusedElectrons

The team used advanced AI to sort patents by SDGs, create summaries and explain why each SDG tag was chosen. Their system offers a user-friendly interface to view trends, applicant data and insights. It also allows users to explore the data using a chat bot.

LIPCY-UP

The team developed a system that uses a set of rules built from existing patent categories to classify patents based on their connection to the UN SDGs. They used advanced AI to help create this rule system and to build a decision tree that analyses the text of each patent to determine whether it has a positive or negative impact on society or the environment.

PatentEmbedders

The five-person team coded a large language model that understands European legislation corpus and further trained it to match pollution abatement techniques using a human-validated dataset. They added a second AI model to improve the accuracy of the results, and included a chatbot that helps users ask questions and receive further insight.

SDG-Concepts

The duo focused their solution on the automatic generation of a silver standard for training and evaluating SDG classifiers. Their method compares rankings instead of using complex math comparisons and they trained their AI to handle multiple goal labels at once, making it more flexible and accurate.

Suma

The two-person team built an AI system made up of five tools, one of which uses a custom model to connect patent categories to SDGs. They used traditional machine learning methods, added extra data for better context and used similarity checks to improve accuracy and reliability.

Vasileios Ntarlagiannis

Vasileios used a group of different AI models to identify which SDGs a patent relates to, based on the text and classification codes. These models work together using a method that helps the system determine how confident it is in its prediction.

Join us for the online prize ceremony on 3 July 

Eager to see who will claim the top spot? Register using the link below to attend the online CodeFest Spring 2025 prize ceremony on 3 July 2025 at 14.40 hrs CEST, as part of the PATLIB conference.  

About CodeFest Spring 2025

This third edition of CodeFest challenged residents of EPO member states aged 18 and over to explore how automated systems for classifying patent data can support the UN SDGs. Developers and data scientists were invited to build tools to help researchers, policymakers, businesses and inventors unlock the value of patent data to accelerate sustainability-focused innovation.

Sustainability is a core focus of the EPO’s Strategic Plan 2028, and CodeFest Spring 2025 reflects this strong commitment to enhancing both the accessibility and strategic use of patent data to support global sustainable development. For more information, please visit the CodeFest Spring 2025 page.