Research Associate at iPrism
Creator & Developer iflora.gr01 Technical Profile
Multidisciplinary researcher with a strong foundation in AI and data science. Experienced in software development, and hardware prototyping, with contributions to international projects and peer-reviewed publications. Skilled in problem-solving, cross disciplinary collaboration, and applying innovative technologies to real world challenges. Passionate about continuous learning, with interests spanning across many fields such as meteorology, and digital health. modus vivendi • modus operandi
Research on distributed data migration and allocation strategies at the edge using graph clustering techniques. Presented at the 27th IEEE International Conference on Mobile Data Management (MDM), Athens, Greece.
Development of a Deep Learning framework to accurately forecast battery health and degradation, providing a real-time decision support system for optimizing the safety and efficiency of electric vehicles and large-scale energy storage. In colaboration with Sunlight Group.
Published research on optimizing data selection in edge computing environments based on task requirements. Novel approach to improve computational efficiency and reduce latency in distributed pervasive systems.
Contributing to an EU-wide initiative enhancing healthcare delivery during cross-border emergencies and routine care utilizing AI-enabled technologies. Focused on IoT integration, patient-centric design, and system resilience.
Developing models to optimize spatial arrangement and coordination of autonomous swarms for improved performance in stochastic environments. Focus on multi-agent coordination and formation control.
Internal lab hub for documenting and managing server resources, operational procedures, and shared tools. It supports two primary compute systems: an AMD server for internal development, datasets, and project workloads, and an NVIDIA server for GPU-intensive Docker experiments and machine learning.
The GitHub Pages/Vite documentation portal is protected by GitHub and Supabase authentication. It organizes server notes, service handbooks, backup strategy, logging and security guidance, plus shared infrastructure services.
Reinforcement learning project focused on adaptive swarm navigation, exploring how autonomous agents can learn coordinated movement and navigation strategies in dynamic environments.
Interactive simulation of flocking behavior using the classic Boids algorithm. Explore emergent swarm dynamics with real-time visualization of separation, alignment, and cohesion rules. A fun way to experiment with autonomous agent behavior and swarm intelligence.
Modular network scanning tool enhanced with artificial intelligence. Features multiple scan profiles (comprehensive, fast, intensive), configurable strategies including SYN stealth, TCP connect, service enumeration, and OS detection. Integrates with Ollama for AI-powered deep search capabilities.
Interactive web application for playing against a model-less agent using Monte Carlo technique. Watch Carlos(the bot) learn optimal Blackjack strategy through configurable training episodes, from quick 10K runs to thorough 500K sessions. Can you win?
A viticultrure and wine center faces some major problems when trying to optimize water and fertilizers management. The repo contains the simulation of a decision support system (DSS) that will propose optimal amounts of water and fertilizers per culture zone, based on previous data and weather forecast.
Unpaired image-to-image translation for face aging and de-aging using CycleGAN. Trained on UTKFace, the four-network system (two generators, two PatchGAN discriminators) learns aging style purely from domain statistics, preserving identity via cycle-consistency and identity losses.
Sequence-to-one LSTM model for monthly temperature forecasting using real meteorological data from Karditsa, Greece, fetched via the OpenMeteo API. Explores and compares multiple LSTM architectures and hyperparameter configurations.
Anomaly detection system for credit card fraud using an autoencoder trained exclusively on legitimate transactions. Flags transactions whose reconstruction error exceeds an auto-tuned threshold, achieving 0.94 ROC-AUC and 80% recall on the Kaggle ULB dataset.
Custom Sequential CNN for multi-class car brand classification. Features a two-phase training strategy, stratified data splitting, class balancing via augmentation, and a named feature-vector head that can be repurposed for similarity search.
Automatic image description pipeline combining a CNN encoder with an LSTM decoder. Trained on Flickr8k and Flickr30k datasets, the model generates natural-language captions by attending to visual features extracted from input images.
iPrism - Contributing to cutting-edge research in distributed systems, edge computing, and AI-enabled technologies for healthcare and pervasive computing environments.
Postgraduate study program. Currently pursuing.
University of Thessaly. Specializing in distributed systems, IoT, machine learning, and telecommunications.
HEAL-Link Consortium by Elsevier. Training on AI-powered research tools and trusted content discovery.
Mathworks certification covering fundamentals of deep learning for image classification using MATLAB software.
The Open University. Covered global meteorological observation methods using satellites, aircraft, balloons, and surface-based sensors.
EU Regulation 2019/947 certification for operating unmanned aircraft systems in the UAS Open category.
Faculty of Medicine, University of Thessaly. Exploring the role of synthetic biology in modern farming.
Department of Informatics with Applications on Bio-medicine, University of Thessaly.
Online WILEY Research Training Program focused on academic research methodologies.
Training on technological tools and specialized applications in information systems security.
Studies in Economics and Informatics. General Grade: 19.7/20
1st LTTA - European educational mobility program.
Certification of PC knowledge in document processing, spreadsheets, and internet services.
