Complex Data Research Group
We find patterns in data and mine information from complexity.
About Us
We build search and analysis methods for complex data: high-dimensional embeddings, protein structures, mass spectra and human motion. Our work covers index design, with learned and graph-based vector indexes, and public search services for life scientists, such as AlphaFind and DreaMS Search. We are part of the DISA laboratory at the Faculty of Informatics, Masaryk University in Brno, and work closely with CERIT-SC. Our code is on GitHub.
Learned and graph-based vector indexes.
We replace hand-tuned partitioning with learned models and study where approximate search loses recall, on dynamic data and on inner-product workloads from LLMs. See LMI at 100M+ objects (SISAP 2024), DaWaK 2025, CIKM 2026.
Search for life-science data.
AlphaFind searches the 240M+ predicted structures in AlphaFold DB by structural similarity. DreaMS Search applies the same approach to mass spectra. Key papers: NAR 2024, NAR 2026.
Clustering and index organization.
Semi-supervised k-medoids (SISAP 2025), metric hulls (SISAP 2021 best student paper), and cluster rebalancing for R-trees (SISAP 2026).
doc. RNDr. Vlastislav Dohnal, Ph.D.
Group co-lead
Vlasta is an associate professor at the Faculty of Informatics, Masaryk University, and a founding member of the Laboratory of Data Intensive Systems and Applications (DISA). His long-standing research covers data management of unstructured data for content-based retrieval and similarity analytics. He has co-authored more than 40 research publications and a seminal book on similarity searching in metric spaces.
RNDr. Matej Antol, Ph.D.
Group co-lead
Matej is executive director of the national research infrastructure CERIT-SC, where he is responsible for integrating the national e-infrastructure e-INFRA CZ. He is also one of the national leaders of Open Science and EOSC implementation in the Czech Republic. His research ranges from index structures and query processing in metric similarity search to research data management and life-science data. Outside science, he plays sax and guitar in a garage band.
RNDr. David Procházka
David works on learned vector indexing and approximate nearest neighbor search in high-dimensional spaces. His Ph.D. studies how learned indexes handle insertions, deletions and data drift. He received the SISAP 2021 Best Student Paper award with Miriama Jánošová, scaled the Learned Metric Index to 100M+ objects, and has a CIKM '26 paper on graph ANN search for attention-derived workloads. He is the tech lead of DreaMS Search and supervises bachelor's and master's theses.
Mgr. Emma Sommerová
Emma is a Ph.D. student working on explainable learned indexes, which she develops as general methods and tests on DreaMS Search, a semantic search platform for hundreds of millions of mass spectra. There she links spectra to experimental metadata from GNPS and ReDU and studies when data from different studies can be indexed together. Her master's thesis built a temporally dynamic learned index, and she co-authored the SISAP 2026 Indexing Challenge paper.
RNDr. Terézia Slanináková, Ph.D.
Terka's Ph.D. research applies machine learning, specifically learned indexing, to fast search over complex data. She is also a researcher in Tom Rebok's group at CERIT-SC, where she applies machine learning to real-world problems and co-leads a project building a national platform for geospatial data analysis. She cares about reproducible research, open source software, and leading bachelor's and master's students. Outside research, she enjoys almost any outdoor sport, meeting friends, and cooking.
Mgr. et Mgr. Jaroslav Oľha, Ph.D.
Jaroslav holds master's degrees in computer science and biochemistry and a 2025 Ph.D. from FI MUNI on data-driven dynamic autotuning. He has published in high-performance computing, similarity searching, and computational chemistry, and his research now focuses on organizing and analyzing complex data, particularly from the life sciences. Outside work, he spends time with his mischievous toddler, plays guitar and bass in a few bands, paints miniatures, plays board games, and runs the occasional marathon.
RNDr. Miriama Jánošová, Ph.D.
Miriama's Ph.D. research focuses on analytics for unstructured data, especially suitable representations of metric regions and their applicability. She also works on motion capture data for rehabilitation therapy and designs exercises for the students' Data Warehousing project. Outside her studies, she is a software engineer at Ataccama, where she develops parsers for SQL-like languages and extracts data lineage from BI reporting tools. In her free time, she trains her dog Brutus, does pilates, and cooks exotic dishes.
Mgr. Samuel Gorta
Samuel is a Ph.D. student supervised by Vlastislav Dohnal, working on molecular dynamics. As a researcher at CERIT-SC, he leads ExTraMD, a CESNET-funded service that extends molecular dynamics simulations by learning the system's behavior in a low-dimensional latent space, so a short input trajectory yields a much longer one. He also contributes to a tool for semantic search in molecular dynamics data. His master's thesis built exponential models of experimental MRI data.
Bc. Boris Lukačovič
Computer Engineering student with a strong foundation in developing software solutions for complex multidisciplinary projects at the intersection of technology and life sciences.
Bc. Tomáš Drkoš
Master's student in Artificial Intelligence and Data Processing, specializing in large-scale data processing and analysis, whose bachelor's thesis studied the quantization of auto-encoded human motion features.
Mgr. Adrián Rošinec
Adrián is a Ph.D. student working on the management of scientific data. He has hands-on experience with research infrastructures, from data storage and processing to authentication and authorization (AAI) and cloud computing.
Bc. Jakub Čillík
Starting his career journey in computer science, Jakub thrives on leveraging his extroverted nature to uncover nuanced narratives hidden in data, remolding it into tangible and expedient solutions.
In our team, we strive to develop cooperation with other research teams from both Czechia and abroad. Our most prominent partners are:
Information Systems and Data Mining group, Kiel University, Germany
With Prof. Dr. Peer Kröger and his group we study indexing and (reverse) kNN retrieval, most recently cluster rebalancing for R-tree bulk loading (SISAP 2026).
IT University of Copenhagen, Denmark
With Martin Aumüller we study why graph-based ANN search fails on inner-product workloads from LLMs and how to recover recall (CIKM '26).
Biological Data Management and Analysis Core Facility of the Central European Institute of Technology
Together, we work on the application of learned indexing to searching in protein structures produced by AlphaFold.
Institute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences
The Tomáš Pluskal Group developed DreaMS, a foundation model for tandem mass spectra. Together we build DreaMS Search, a semantic search engine over large spectral collections.
Czech Institute of Informatics, Robotics and Cybernetics at the Czech Technical University in Prague
Josef Šivic's Intelligent Machine Perception group co-developed the DreaMS model with IOCB Prague. Together we work on DreaMS Search, combining their representation learning with our similarity search and indexing.
e-INFRA CZ / CERIT-SC, Masaryk University
Engineers from e-INFRA CZ help us with fine-tuning our algorithms and running our experiments, often requiring non-trivial know-how and computing power.
Loading map…
Selected Publications
2022
-
Learned Indexing in Proteins: Substituting Complex Distance Calculations with Embedding and Clustering Techniques
Similarity Search and Applications, 15th International Conference, SISAP 2022, Bologna, Italy, October 5–7, 2022, Proceedings, year: 2022, DOI
2021
-
Data-driven Learned Metric Index: an Unsupervised Approach
14th International Conference on Similarity Search and Applications (SISAP 2021), year: 2021, DOI
-
Learned metric index - proposition of learned indexing for unstructured data
Information Systems, year: 2021, volume: 100, edition: 101774, DOI
-
Metric hull as similarity-aware operator for representing unstructured data
Pattern Recognition Letters, year: 2021, volume: 149, edition: September 2021, DOI
-
Organizing Similarity Spaces using Metric Hulls
14th International Conference on Similarity Search and Applications (SISAP 2021), year: 2021, DOI
Interested in Our Research? Join Us!
We are searching for new colleagues for various positions who would work with us on exciting projects, develop unique software and solve unconventional problems. In case of interest, contact is at dohnal(at)fi.muni.cz or antol(at)muni.cz.