At T.S.I., I'm continuing my work under Parasecurity Group, spearheading our research on LLMs. I am developing an efficient Mixture-of-Experts LLM with alternative attention schemes and optimizations for efficiency.
At ICS-FORTH, I'm working with the Parasecurity Group. I have developed algorithms that perform various tasks, including sentiment analysis, named entity recognition, and topic modeling. I have also explored deep learning techniques, such as recurrent neural networks and convolutional neural networks, to develop models that can learn text data representations and make accurate predictions.
This study analyzes how Twitter's suspension mechanism relates to policy violations amidst the Russo-Ukrainian War discourse, using a dataset of 107.7M tweets. Results reveal scams, spam, and ads exploiting the conflict's trending topics, with machine learning explaining suspension triggers.
Using a multilingual dataset of 9M users discussing the Russo-Ukrainian war, this study detects Twitter bot accounts spreading misinformation. We used 343K bot and 8M normal accounts, enhancing bot detection using an XGBoost model. Compared to Botometer, our method improved ROC-AUC score by 11%.
Our study uses an 89M tweet dataset to examine Twitter discourse on the Russo-Ukrainian War, employing zero-shot learning due to diverse languages and lack of labels. It explores sentiment, topic narratives, toxicity levels, and military-related communication. Findings shed light on social media's role in conflicts.
Our team analyzes 2023 Greek election-related tweets using Facebook's BART zero-shot classification model for sentiment analysis. This powerful tool discerns public mood and insights into Greece's political landscape without requiring pre-labeled data, offering a timely, insightful analysis of the election.
I am John Lamprou, I'm an MSc Candidate at ECE TUC and an R&D Engineer at Telecommunication Systems Research Institute (T.S.I.), Technical University of Crete, part of the Parasecurity Group.
Passionate and independent at heart, I am a Machine Learning/NLP Engineer with a zest for pushing the boundaries of Large Language Models (LLMs). My journey is marked by a relentless quest to reimagine and refine the architectures of LLMs, ensuring they are not just groundbreaking in theory but also efficient and scalable in the real world. At the core of my work is a deep commitment to innovation—whether it's enhancing model efficiency, reducing computational demands, or translating complex research into user-friendly applications. With a proven knack for navigating the entire research and development continuum solo, I pride myself on my ability to turn visionary ideas into tangible, impactful technologies.