Research Interests

My research lies at the intersection of Generative AI, Deep Learning, and Natural Language Processing (NLP). I am particularly interested in applications across healthcare, low-resource languages, and increasingly, cybersecurity.

With a Ph.D. in Computer Engineering and three postdoctoral appointments, I have led or contributed to a wide range of AI and NLP projects, and supervised over 30 master's theses. My current focus centers on how Generative AI can drive innovation not only in language understanding and generation but also in enhancing the resilience and interpretability of intelligent systems.

Previous Research Contributions

Biomedical Named Entity Recognition

During my doctoral research, I tackled the challenges of Named Entity Recognition (NER) in the biomedical domain. This area is complicated by the frequent use of synonyms, abbreviations, and ever-evolving terminology. I developed ensemble learning methods and applied metaheuristic algorithms like Particle Swarm Optimization (PSO) to improve recognition performance in chemical and biomedical texts.

Automated ICD Coding

At KU Leuven, I investigated the automation of International Classification of Diseases (ICD) coding using deep learning. By classifying clinical reports into ICD categories, this work aimed to streamline medical documentation and reduce manual coding efforts in healthcare institutions.

Causal Relationship Extraction

Also at KU Leuven, I explored methods for extracting causal relationships from biomedical literature. I evaluated models such as Multiview CNNs, attention-based BiLSTM architectures, and modern word embeddings like ELMo and BioBERT. A significant part of this work involved mitigating class imbalance to enhance performance.

Word Sense Induction (WSI)

At the University of Zagreb, I designed a novel Leader-Follower clustering algorithm to tackle Word Sense Induction, focusing on identifying word meanings using automatically generated lexical substitutes.

Current Focus and Future Plans

My current work revolves around advancing Generative AI techniques for various real-world challenges. This includes Visual Commonsense Reasoning (VCR), in-context learning, and evaluating summarization quality using Large Language Models (LLMs) without reference summaries.

1. Generative AI for Cybersecurity Applications

I am investigating how Generative AI can be used to strengthen cybersecurity through:

• Simulating cyber threats such as phishing and malware for red teaming exercises
• Detecting anomalies via generative log modeling
• Building autonomous agents for threat response and mitigation
• Deploying AI-powered honeypots for cyber deception

2. Explainable Multi-Modal Language Models

As Multi-Modal Language Models (MLLMs) grow more complex, interpretability becomes critical. I aim to build explainable frameworks to improve transparency and fairness, particularly in sensitive domains such as healthcare and digital forensics.

3. Agentic AI and Intelligent Process Automation

I am exploring how agent-based AI systems can integrate with Intelligent Process Automation (IPA) to streamline operations in sectors like healthcare, finance, and customer support.

4. Hyper-Personalization and Emotionally Resonant Content

My research also targets the development of AI systems that can produce hyper-personalized and emotionally intelligent content. By leveraging multimodal generation (text, image, video, speech), these systems can offer highly engaging user experiences tailored to individual behaviors and preferences.

5. Generative AI for Low-Resource Languages and Multimodal Reasoning

Many linguistic communities are underrepresented in AI research. I aim to develop efficient generative models for low-resource languages through transfer learning, semi-supervised approaches, and robust model architectures. I also plan to enhance multimodal reasoning in Vision-Language Models (VLMs) to enable better understanding of visual-textual interactions.