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. First things first, some personal promo! 😎
Check out my image tiling library plakakia and my latest streamlit app demos for #image and #signal for processing. Let me know if they're sleeping :)
TL;DR: A computational scientist specializing in AI & ML, combining backgrounds in Computer Science, Machine Learning, and Bioscience Engineering. With hands-on experience in analyzing neurophysiological data using Neural Networks and developing AI software solutions in a startup, I served as a Postdoc Researcher at KU Leuven's MeBioS Biophotonics Group, continuing after PhD tenure, overseeing insect-monitoring and agri-food projects, mentoring PhD researchers, MSc/BSc students, managing the lab's data and software, plus fostering AI adoption across diverse present and potential future projects. 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.
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.
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 which ran on #IoT (e.g., RaspberryPi) devices, Linux/Windows desktops, and the cloud (#AWS). My latest achievement is 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 incorporates 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:
🧬 orc-id
🔬 Google Scholar
📖 ResearchGate
🌐 Stay connected through the following social media channels:
📲 X/Twitter
📲 LinkedIn
📲 GitHub