Dr. Ceca Kraišniković

Research Roles

Deep Learning & Computational Neuroscience & Energy-efficient computing

PhD Research: University Assistant & Scientific Project Associate @ Graz University of Technology

Worked at the Institute of Machine Learning and Neural Computation (IML) with Dr. Robert Legenstein. Focused on modeling learning and computation in biological and artificial neural systems, with applications to energy-efficient neuromorphic hardware. Investigated spiking neural networks with memory mechanisms for solving cognitive tasks, including symbol and arithmetic sequence manipulation, working memory dynamics, and information coding. Additionally explored training paradigms for sparse neural networks with faulty or noisy memristive synapses.

Large Language Models for pathology text reports

Postdoctoral Researcher @ Medical University of Graz

Conducted AI research on medical data analysis at the Diagnostic and Research Institute of Pathology. Applied large language models (LLMs) to pathology text reports, focusing on continued pre-training, fine-tuning, and knowledge representation in embedding spaces, and interpretability of the fine-tuned models for tumor classification based on text reports. Developed methods for annotating unstructured medical text to facilitate integration of text and image data, managing large medical databases, and implementing a Python Extract-Transform-Load (ETL) interface to support research data access.

Machine Learning for Signal and Medical Data Processing

Project Associate @ Human Research Institute of Health Technology and Prevention Research, Weiz, Austria

Improved detection algorithms for physiological signals (ECG and pulse wave) through algorithm search, runtime evaluation, parameter optimization, and statistical performance comparison. Implemented the best solution in C on an embedded ARM Cortex-M0 platform. Developed methods for synchronizing and comparing ECG signals across different measuring devices.

Publications, Posters & Talks; Theses, Research Projects

Publications

  1. Kraišniković, C., Harb, R., Plass, M., Al Zoughbi, W., Holzinger, A., & Müller, H. (2025). Fine-tuning language model embeddings to reveal domain knowledge: An explainable artificial intelligence perspective on medical decision making. Engineering Applications of Artificial Intelligence, 139, 109561. (URL)
  2. Kraišniković, C., Stathopoulos, S., Prodromakis, T., & Legenstein, R. (2023, June). Fault Pruning: Robust Training of Neural Networks with Memristive Weights. In Unconventional Computation and Natural Computation: 20th International Conference, UCNC 2023, Jacksonville, FL, USA, March 13–17, 2023, Proceedings (pp. 124-139). Cham: Springer Nature Switzerland. (URL)
  3. Kraišniković, C., Maass, W., & Legenstein, R. (2021). Spike-Based Symbolic Computations on Bit Strings and Numbers. In Neuro-Symbolic Artificial Intelligence: The State of the Art (pp. 214-234). IOS Press. (URL)
  4. Salaj*, D., Subramoney*, A., Kraisnikovic*, C., Bellec, G., Legenstein, R., & Maass, W. (2021). Spike frequency adaptation supports network computations on temporally dispersed information. Elife, 10, e65459. (URL)

Posters

  1. Steinwender, L., Beck, P. G., Hambleton, K., Kraisnikovic, C., Stadlober-Temmer, M., & Hanslmeier, A. (2022, July). Machine learning classification of RR Lyrae stars observed by TESS. In TASC6/KASC13 Workshop. (URL)
  2. Kraisnikovic, C., Maass, W., & Legenstein, R. (2021, September). Spike-based symbolic computations on symbol sequences. In Bernstein Conference 2021. (URL)
  3. Subramoney, A., Kraisnikovic, C., Salaj, D., Bellec, G., Legenstein, R., & Maass, W. (2020, September). Spike-frequency adaptation contributes long short-term memory to networks of spiking neurons. In Bernstein Conference 2020. (URL)

Talks

  1. Barcelona Supercomputing Center (NLP for biomed research group) - Seminar presentation (online), October 2025: Fine-tuning language model embeddings to reveal pathology domain knowledge (pdf)
  2. Unconventional Computation and Natural Computation (UCNC) Conference 2023: Fault Pruning: Robust Training of Neural Networks with Memristive Weights. (pdf)
  3. Infineon Winter School 2023: Learning in biological and artificial neural networks - for understanding the brain and developing energy-efficient hardware implementations. (pdf)
  4. Bernstein Conference 2021: Spike-based symbolic computations on symbol sequences. (pdf)
  5. Neuromatch (Un)Conference 2020: Spike-frequency adaptation enables cognitive computations. (pdf)
  6. Bernstein Conference 2020: Spike-frequency adaptation contributes long short-term memory to networks of spiking neurons. (pdf)
  7. Styrian Brain Research Initiative - Networking Event @ Graz University of Technology, 2019: (Event info) (pdf)

Theses

European Union (EU) Research Projects