Performance, scalability and energy-aware HPC
I study how scientific applications behave on modern CPU and GPU architectures: profiling them, modelling the trade-off between execution time and energy, and optimising them for production systems. Recent work covers CFD codes on AMD EPYC clusters, multi-card GPU machine-learning jobs and Python-based applications on HPC systems.
Selected publications
- Data-Driven Prediction of Optimal MPI Rank Placement and Node Count for Memory-Bound CFD on AMD EPYC Clusters (arXiv preprint, 2026)
- Evaluating AMD EPYC CPU architectures on CFD applications (Future Generation Computer Systems, 2025)
- Prediction model of performance-energy trade-off for CFD codes on AMD-based cluster (Future Generation Computer Systems, 2025)
- Profiling and Optimization of Multicard GPU Machine Learning Jobs (Concurrency and Computation: Practice and Experience, 2025)
- Profiling and optimization of Python-based social sciences applications on HPC systems by means of task and data parallelism (Future Generation Computer Systems, 2023)
Software: Quare, a framework for analysing application performance-profiling data. Project deliverables on benchmarking and optimisation: HiDALGO2 and HiDALGO.
Go to top of the page.
Digital twins and large-scale environmental simulation
I work on coupling simulation models with data analytics so that environmental and global-challenge applications can run at scale and be operated as services: urban air quality, building energy, renewable energy, wildfire propagation and water-borne material transport. Since 2024 I also co-organise the AGCA workshop series at PPAM, which focuses on advancements of such applications.
Selected publications
- Interpretable Patch-Based Deep Learning for Wildfire Spread Prediction from Ensemble Simulations (arXiv preprint, 2026)
- Large-scale parallelization of Human Migration Simulation (IEEE Transactions on Computational Social Systems, 2024)
- P-FLEE: An Efficient Parallel Algorithm for Simulating Human Migration (IEEE IPDPS Workshops, 2021)
Projects: HiDALGO2 (principal investigator), HiDALGO (quality manager, work package leader), CoeGSS (work package leader); deliverables are listed under Reports. Workshops: AGCA @ PPAM.
Go to top of the page.
HPC–HPDA–AI workflows and data management
Beyond raw performance, I work on how data, AI and HPC fit together in practice: allocating image-recognition and LLM tasks across multi-GPU systems, fusing open geodata for analysis on HPC, and moving datasets between data portals and HPC storage so that research data stays citable and reproducible.
Selected publications
- Efficient Allocation of Image Recognition and LLM Tasks on Multi-GPU System (Lecture Notes in Computer Science, Springer, 2025)
Software: Qumasc, CKAN2HPC-Client and CKAN-Zenodo-Exporter. Project: MARVEL (HPC infrastructure and resource management for audio-visual data analytics; deliverable D5.3).
Go to top of the page.
Scheduling theory: divisible load
My doctoral work (Poznań University of Technology, 2006) concerned scheduling of divisible loads, that is, arbitrarily splittable workloads, on parallel and distributed systems with limited memory. The theory remains the basis for the scheduling tools I build today.
Selected publications
- Experimental study of scheduling with memory constraints using hybrid methods (Journal of Computational and Applied Mathematics, 2009)
- Scheduling multiple divisible loads in homogeneous star systems (Journal of Scheduling, 2008)
- Multi-installment divisible load processing in heterogeneous distributed systems (Concurrency and Computation: Practice and Experience, 2007)
- Scheduling multiple divisible loads (International Journal of High Performance Computing Applications, 2006)
Software: DLT Studio, the solver library behind the DLT Portal (DOI 10.5281/zenodo.23299701).
Go to top of the page.
Research infrastructures for science
For over two decades I have helped build the infrastructures scientists use: virtual laboratories giving remote and interactive access to instruments in grid environments, e-infrastructure policy work, and digital repositories of natural history collections with analytical tools and data exchange with GBIF.
Selected publications
- Determination of the relativistic value of museum natural collections (Journal of Natural History, 2023)
- Open Access to the Digital Biodiversity Database: A Comprehensive Functional Model of the Natural History Collections (Diversity, 2022)
- The Adam Mickiewicz University Nature Collections IT system (AMUNATCOLL): metadata structure, database and operational procedures (Biodiversity: Research and Conservation, 2022)
- Virtual Laboratory as a Remote and Interactive Access to the Scientific Instrumentation Embedded in Grid Environment (IEEE International Conference on e-Science and Grid Computing, 2006)
- Dynamic Measurement Scenarios in the Virtual Laboratory system (IEEE/ACM International Workshop on Grid Computing, 2004)
- General Conception of the Virtual Laboratory (ICCS, Lecture Notes in Computer Science, 2004)
Projects: DORII (work package leader), RINGrid (deputy principal investigator), EXPReS, e-IRGSP5 (work package leader), AMUNATCOLL (technical coordinator). Software: AMUNATCOLL IT, CollMan.
White papers and reports for the e-Infrastructure Reflection Group (e-IRG)
- Report on the retrieval, provision and analysis of cost-related policy information (e-IRG report, February 2019)
- Summary report – Report on the retrieval, provision and analysis of policy and cost-related information / indicators (e-IRG summary report, February 2019)
- Report on the Retrieval, Provision and Analysis of Policy and Cost-Related Information / Indicators (e-IRG report, January 2019)
- Evaluation of e-Infrastructures and the development of related Key Performance Indicators (e-IRG report, March 2017)
- e-IRG Roadmap 2016 (e-IRG white paper, December 2016)
Go to top of the page.