When
April 16 — June 15, 2026
Prize pool
$7000
Format
Kaggle challenge with final event in Kyiv
Participation
Free
About:
While the government is actively going digital, we still lack simple and open tools for working with Ukrainian handwritten documents. Most existing solutions are either closed or assembled from separate components and do not scale well. As a result, this slows down the launch and development of services like E-license and other digital public services.
That’s why AI HOUSE, together with the Ministry of Economy, the Ukrainian Catholic University and with the support of the Ministry of Digital Transformation, is launching an AI challenge where participants will build AI solutions to recognize Ukrainian handwritten documents: applications, certificates, logs, signatures, stamps, and archival materials, for real-world processes and tasks.
During the challenge, you won’t just train models on a dataset. You’ll help create tools that can be implemented in the work of the Ministry of Economy and the Ministry of Digital Transformation, simplify document processing, and speed up the delivery of public services.
Who is it for:
- AI/ML engineers, data scientists, researchers
- Students of technical fields working with Computer Vision, OCR/HTR, image and text processing
- Ukrainian and international startups, R&D teams, and university labs interested in practical AI applications for working with documents and public data
The challenge will take place in two stages:
- Online stage [Kaggle]: An open competition with no separate registration. Participants develop handwritten text recognition models, and the results are automatically evaluated on the platform.
- Final [offline in Kyiv]: Top teams based on the results of the online stage will present their solutions live: demonstrating model performance, their approach to reproducibility, and real results on the data.
What you'll gain:
- A $7,000 prize pool
- A chance to contribute to the digitalization of government processes and the creation of open infrastructure for document processing
- The opportunity to test your solutions on real use cases of government services, including eDozvil
- Mentorship support from experts in computer vision and OCR/HTR
- Access to a unique open dataset of Ukrainian handwritten documents from different domains and handwriting styles (state archives, universities, NSTU, and other sources)
- AWS credits for training and testing models
- The possibility of further implementation of solutions in the public sector together with the project’s partners, including the Ministry of Economy of Ukraine
Registration and participation:
— Access to the Kaggle task opens on April 16 and runs until June 15. On July 4 we will hold the final offline event;
— You can participate in teams of 2–5 people, individually, or join existing teams;
— The number of participants is unlimited.
Experts who will support the teams
- ML Engineer with over 8 years of commercial experience in developing ML solutions, including LLM applications, recommendation systems, search, computer vision, and NLP.
- Has deep hands-on expertise in implementing MLOps and LLMOps and advises teams on designing scalable ML systems.
- Works as the ML Lead of the national educational platform Mriya, where is responsible for AI functionality, and also serves as an Advisor to the Ministry of Economy of Ukraine, supporting the country’s digital transformation initiatives.
- Researcher and engineer with over a decade of experience in NLP, LLMs, multimodal models, and building AI systems.
- Holds a PhD in Computer Science, leads the NLP Lab at the Kyiv School of Economics, and works as a Data Scientist at Visa, where develops and deploys AI solutions in complex enterprise environments. Leads interdisciplinary teams, bridging fundamental research with practical applications.
- Teaches at KPI, is a co-organizer of the UNLP conference, and actively contributes to the development of the Ukrainian AI community.
- For over ten years, she has led a team of computational linguists at Superhuman, working on algorithms for error correction and text improvement.
- She has deep expertise in NLP—from syntactic parsing to sentiment analysis and fact extraction—and pays particular attention to developing tools for processing the Ukrainian language.
- Mariiana is an organizer of the UNLP conference, a co-lead of the lang-uk initiative, and a member of the Expert Committee on AI Development at the Ministry of Digital Transformation of Ukraine.
- Data Scientist specializing in NLP, Generative AI, Agentic AI, and Responsible AI.
- Leads AI research teams, works on LLM evaluation, multimodal RAG systems, multi-agent architectures, synthetic data generation, and code modernization.
- Actively advances the field of ethical AI use, creates educational materials on NLP, and shares expertise through lectures and public talks.
- Data Scientist with experience working on projects in the public sector, healthcare, retail, and finance—from data analytics to recommendation systems.
- Actively contributes to open-source initiatives, developing speech synthesis and speech recognition models with a focus on low-resource languages, including Ukrainian and Crimean Tatar.
- Project Lead of Lapa LLM and a co-creator of one of the most efficient Ukrainian large language models within the Made Ukrainian NLP Great initiative.
- Works on developing Ukrainian language models and post-training large language models within international open-source initiatives.
- Actively contributes to building safe multilingual AI systems with organizations such as ETH Zurich, Hugging Face, and Cohere Labs, promoting Ukrainian language in the global AI ecosystem.
- Focuses on LLM capabilities and safety across languages, with hands-on experience across the full development cycle: from large-scale training on GPU clusters to adapting models for real-world applications.
Event organized by:
Event organizers:
With support from:
General partners:
Technical partners:
Data partners:
FAQ
-
How to participate?
You just need a Kaggle account, go to the challenge page, and join the competition. You can participate with your own team [2–5 members], individually, or join existing teams. See the full list of rules and participation terms at the link.
-
Is participation paid?
No, participation is free — and teams will compete for a $7,000 total prize pool.
-
Can I participate remotely?
Yes, the entire challenge takes place online. However, the final stage for the challenge finalists will be held offline in Kyiv.
-
When does it start and end?
The challenge starts at the moment the task is published on Kaggle. The online stage lasts until June 15, after which the finalists are determined.
-
What is the language of the event?
Ukrainian. The datasets are in Ukrainian.
-
Still have questions?
Write to: events@aihouse.org.ua