I’m an undergraduate Computer Science student at the Ionian University, Department of Informatics, with a growing academic and research presence in Machine Learning, Natural Language Processing (NLP), and Explainable AI (XAI).

I’m interested in building intelligent systems that are both effective and interpretable: How do we design modular AI systems that use natural language as both input and programming substrate? How can we optimize their performance and transparency using state-of-the-art language models?

To answer these questions, my work explores transformer-based models, semantic retrieval pipelines, and techniques for explainable model behavior. This includes research on discourse and sentiment analysis, systems that extract structured knowledge from public discussions, and tools that enhance interpretability in AI pipelines through visualization and modular design.

I’ve co-authored peer-reviewed research presented at international conferences such as AIAI 2025 at the age of 18, and I’m currently preparing additional work on explainable transformer architectures and scientific document understanding. I aim to pursue a Ph.D. focused on AI/ML, potentially at institutions like ETH Zurich, and my long-term goal is to contribute to foundational advances in explainable, modular AI.

---

Research

My research focuses on two main directions, unified by the goal of building interpretable and modular AI systems that address real-world problems. This work is reflected in published peer-reviewed research and full-stack academic tools.

      

I) Building Reliable & Interpretable AI Systems

I co-developed AI Scanner, a full-stack system for document analysis combining OCR, transformer-based classification, and explainable output generation. My current research investigates:

Natural Language Processing Pipelines for extracting sentiment, topics, and structure from long-form or noisy user-generated content using models like BERT and RoBERTa.

Explainable AI Architectures for deep learning models applied to scientific and policy texts, aimed at improving transparency in model predictions through visualization and interpretation layers.

II) Developing Transformer-Based Systems for Social Discourse Understanding

My paper "Analyzing Public Discourse and Sentiment in Climate Change Discussions" (AIAI 2025) presents a transformer-based pipeline for large-scale social text mining. It explores:

Representation Learning for noisy, multilingual Twitter data to extract coherent, structured insights about global climate discussions.

Semantic Reasoning and document clustering for policy-level decision support in sustainability and environmental contexts.

Additional work is in progress on explainable scientific literature classifiers and document-level segmentation models.

---

Papers

Analyzing Public Discourse and Sentiment in Climate Change Discussions Using Transformer-Based Models
N Roufas, A Mohasseb, I Karamitsos & A Kanavos*
IFIP AIAI 2025 | paper

---

Last Update: July 2025