Here you can find a brief, yet complete, overview of my background. For a summary of links to various online profiles, you can check out my linktree.
Also, feel free to have a look at my image tiling library plakakia
A multidisciplinary computational scientist at the intersection of AI, Machine Learning, and Bioscience Engineering, with a proven track record of translating cutting-edge research into impactful technological solutions. Leveraging advanced expertise in Neural Networks, computer vision, and data-centric AI, I specialize in developing innovative research and applied AI software solutions across domains—from neurophysiological data analysis to environmental monitoring. My professional journey spans academic research, industry applications, and strategic AI implementation, with a consistent focus on driving meaningful technological advancements that address real-world challenges.
I studied Computer Science in the Aristotle University of Thessaloniki (Greece 🇬🇷) earning a solid basis around computing theory. Next, I finished my Master's in Machine Learning at KTH University (Stockholm, Sweden 🇸🇪) specializing in Computational Neuroscience (Spiking Neural Networks). For my thesis work, I simulated a small piece of the neocortex using the NEST simulator in Python to compare various columnar structure types and their activity. My academic journey continued with two years of research in a neurophysiology lab, exploring computational neuroscience. While I did not complete the initial PhD program, I subsequently earned a PhD in Bioscience Engineering, pivoting my research to focus on optical insect identification using artificial intelligence.
As a PhD researcher in the lab of Neurophysiology of KU Leuven for 2 years, I conducted in-depth studies on deep Convolutional Neural Networks and their resemblance to the visual system. My work ([1][2][3][4]) included complex computer vision and regression tasks for predicting biological neuronal activity based on artificial neuron activations of various SOTA CNN models, leading to 4 scientific publications in renowned Neuroscience journals and a poster presentation at VSS conference (Florida, USA), before exiting the programme.
Having developed a passion for #Deep-Learning and its software ecosystem, I wanted to shift my focus from fundamental research to applied AI applications for which I could more clearly gauge their societal impact. Working as a Data Scientist at Faktion in Antwerp, I honed my skills in industry practices such as end-to-end ML pipelines, AI model training, Docker containers, and Cloud components. Notably, my team and I won a hackathon on Activity Recognition in video data, organized by Vinci Energies.
Motivated to pursue more applied research this time, and be closer to home, I returned to Leuven (and KUL) to obtain my #PhD in Bioscience Engineering. My thesis topic was Optical Insect Identification using Artificial Intelligence and focused on 2 distinct insect recognition tracks based on:
- images, using Computer Vision,
- time-series (wingbeats), using Signal Processing.
The main objectives of my research were around data-centric AI and strict model validation to reveal the "true" model performance once deployed in the field. During my PhD I have developed software tools, GUIs (#Streamlit, #Tkinter) and AI models (YOLO, RCNN, 2-stage detectors, ...) which ran on #IoT (e.g., RaspberryPi) devices, Linux/Windows desktops, and the cloud (#AWS). My latest achievement was a Streamlit & #FastAPI server that runs on AWS and serves our image classification model to external companies and collaborating research institutes (examples of device and software: 1, 2). Apart from the API, it incorporated a user-friendly GUI to aid researchers with image annotation tasks.
As a Postdoctoral researcher at MeBioS (KUL), I got involved in multiple projects around AI in insect monitoring or agrifood applications. I guided PhD researchers and built software tools that aided in their research. Being more involved in Hyperspectral Imaging (#HSI) projects, I familiarized myself with SOTA techniques to deal with complex hypercube data using AI. Moreover, I was the research data and software manager for our lab, being responsible on hosting and sharing our software/data using KUL's infrastructure and maintaining our research group's #GitLab (here's its public profile, where you can see some of its content).
Now, I'm taking my expertise to new heights as a remote sensing & AI researcher at Vito. My current role involves classifying the earth's land cover in a reliable and accurate way through the LCFM project of the EU commission (JRC). This important work has real-world applications for environmental conservation, land use planning, and climate change mitigation.
By staying up-to-date with technological advancements, my commitment is to make meaningful contributions to the field of pattern recognition. Let's collaborate to create practical solutions that have a real impact! 🔧
🌱 I’m always interested to learn about how Artificial Intelligence can improve our lives.
💬 Do you want to reach out? Send an email at kalfasyan[at]gmail[dot]com
🔗 Check my linktr.ee
📚 Researcher profiles:
🧬 ORCID
🔬 GOOGLE SCHOLAR
📖 RESEARCHGATE
🌐 Stay connected through the following social media channels:
📲 BLUESKY
📲 LINKEDIN
📲 GITHUB