vasilis@summit:~$ whoami
Machine Learning & AI Engineer
Bachelor's degree in Computer Science at the University of Piraeus with a strong passion for machine learning and AI engineering. I enjoy turning ideas into real-world solutions through code — from training and evaluating deep learning models to building computer vision systems and multi-agent AI tools — and I'm constantly looking for opportunities to expand my knowledge and skills.
02~/about
I'm passionate about coding and love bringing ideas to life through software. Creating useful, impactful applications is what drives me every day — and lately that means machine learning and AI: training models, evaluating them honestly, and turning them into tools people can use.
// off the keyboard: usually somewhere above the treeline
University of Piraeus
ZERONET (Θ. Παπαπασχάλης - Ι. Πεσλής Ο.Ε.)
Managed Microsoft systems including Azure AD, configured and maintained professional IT equipment, and supported network management and troubleshooting.
03~/projects
Mostly machine learning and AI — deep learning, computer vision and agents — plus a few side trails.
A computer-vision search for the world's most elusive striped man. Three detectors — a sliding-window CNN written from scratch (with hand-implemented IoU and non-max suppression), YOLO11n on whole pages, and YOLO11n fine-tuned on native-resolution tiles — compared with scene-level 5-fold cross-validation. The best one powers a themed Streamlit app that ranks the likeliest spots with zoomed-in close-ups.
Waldo in the top 3 guesses on 72% of unseen pages
An experimental comparison of a baseline CNN, EfficientNet-B0 and DenseNet-121 for detecting pneumonia in chest X-rays. Every model was trained five times and reported as mean ± std, with class weighting, threshold optimisation and Grad-CAM heatmaps showing where each network looks.
DenseNet-121: 0.896 ± 0.011 accuracy across 5 runs
Source Code of Pneumonia Detection from Chest X-Rays on GitHubAn intelligent multi-agent system that automatically conducts literature reviews, searches academic papers, and generates professional PDF reports.
Source Code of Research ArXiv Assistant on GitHubA comprehensive machine learning project that predicts Pokemon battle outcomes using statistical analysis and advanced feature engineering techniques.
Source Code of Pokemon Battle Prediction on GitHubFinds YouTube creators in any niche through the YouTube Data API v3, then enriches each channel with ViewStats analytics scraped with Selenium — subscribers, views, 28-day growth and estimated revenue — into a CSV database for creator research and outreach.
Built with @nikosgravos
Source Code of YouTube Creator Scraper on GitHubFridgeChef is a smart Android app that helps users find delicious recipes based on the ingredients in their virtual pantry, with features like ingredient image recognition, step-by-step cooking instructions, and easy recipe sharing via social media.
Source Code of FridgeChef on GitHubFever Dream is a 3D adventure game developed in Unity using C# for Brackeys Game Jam 2025.1. The game follows a lumberjack on his journey to his cave for a nap, where unexpected challenges arise, aligning with the jam's theme, 'Nothing Can Go Wrong'.
Source Code of Fever Dream on GitHubWant to see more of my work?
View All Projects04~/basecamp
Leave a note at basecamp and I'll get back to you soon.