The 4th International Conference on AI in Cybersecurity (ICAIC-2025) was held in Houston on Feb 7, 2025. It was a three-day program at the University of Houston with more than 200 research submissions and participants from 16 countries. The hybrid event brought together academics, engineers, and industry leaders to discuss how artificial intelligence is shaping cybersecurity and beyond.
Among the most visible contributors was Rohit Singh Raja, Associate Director of Quality Engineering at a leading global clinical research organization. Raja chaired a session and stood out as the most published author at this year’s conference, presenting three IEEE research papers that linked healthcare, policy, natural language processing, and big data to the wider challenges of cybersecurity and innovation.
One of Raja’s papers, A Multi-Level NLP Framework for Medical Concept Mapping in Healthcare AI Systems, tackled the messy world of healthcare text. Hospitals and medical records often use different terms for the same conditions or drugs, creating confusion for AI systems. Raja presented a three-layer framework that combines tokenization, knowledge graphs, and semantic analysis to bring consistency to medical data. Tested against a benchmark dataset, it achieved 96.81% accuracy, a result that could help reduce errors in large-scale healthcare systems.
Another paper, Promoting Health through Transformative Innovation Policies and Emerging Technologies, shifted the focus from systems to policy. Using the UK’s mission to apply AI against chronic disease as a case study, Raja showed that AI projects in this area were 37% less common than expected, pointing to a gap between policy goals and research activity. He also found signs of progress: more medical and biotech experts are now working alongside computer scientists, suggesting a move toward cross-disciplinary teams. The study proposed indicators that governments and funders can use to track whether policy promises turn into measurable results.
In his third paper, Time-Varying In-Hospital Mortality Prediction with Apache Spark, Raja explored predictive analytics for hospitals. Instead of predicting risk only at ICU admission, his model updated predictions over the course of a patient’s stay. Built on the MIMIC-III database (a large-scale hospital database) and deployed on Apache Spark for scalability, the system combined clinical notes with standard severity scores. It reached an average accuracy (AUC) of 0.878, but when notes were excluded, performance dropped significantly, showing how valuable unstructured text can be. Raja emphasized that models must not only perform well but also remain interpretable so that clinicians can trust them.
He also argued that technologies should be judged less by novelty and more by measurable improvements in areas like patient safety, efficiency, and compliance.
The conference reinforced this blend of technical depth and real-world applications. Professors Hardik Gohel (University of Houston–Victoria) and Bishwajeet Kumar Pandey (GL Bajaj College, India) oversaw the proceedings, ensuring rigorous review and practical relevance across submissions.
Taken together, Raja’s papers captured the core message of ICAIC-2025: technology can no longer be siloed. Whether it is natural language processing, hospital analytics, or national innovation policy, AI now operates at the crossroads of health, security, and governance. The challenge is not just building models but ensuring they work in high-stakes environments.
As ICAIC-2025 concluded in Houston, the theme was clear. The future of AI in cybersecurity and in healthcare will depend not only on breakthroughs in code but on how well those systems adapt to the realities they are meant to serve.
