Research & Development Projects
Pushing the Limits of Correlative Nanoscopy with Generative Artificial Intelligence
Role: Researcher | Org: Zoi Data
Funding Program: Bilateral Cooperation Program with Romania (MCID)
This project aims to enhance correlative nanoscopy techniques by integrating generative artificial intelligence models for image reconstruction, enhancement, and cross-modality translation. Correlative nanoscopy combines multiple imaging modalities (e.g., fluorescence and electron microscopy) to provide detailed structural and functional information at the nanoscale. However, image acquisition and alignment across modalities remain challenging. By leveraging the power of generative AI, such as GANs and diffusion models, the project seeks to bridge resolution gaps, reduce acquisition time, and improve the interpretability of nanoscopic datasets. As part of a Turkish-Romanian collaboration, Zoi Data contributes expertise in AI model development and computational image analysis.
BB-REBUS – Brain-Body factoRs mediating altEred Bodily representations
Role: Researcher | Org: Zoi Data
Funding Program: European Commission – JTC2024
BB-REBUS is a transdisciplinary European research project aiming to investigate the neural and bodily mechanisms underlying distorted bodily representations observed in various pathological conditions, such as chronic pain, eating disorders, and neurological syndromes. By integrating neuroscience, clinical research, and computational modeling, the project seeks to identify common brain-body interaction factors contributing to altered self-perception. As the Turkish partner, Zoi Data is responsible for developing advanced machine learning solutions for analyzing neurophysiological and behavioral data.
Early Diagnosis of Cardiac Ischemia: Melanin-Enhanced Biosensor
Role: Coordinator | Org: İzmir Democracy University
This project aims to develop a novel, paper-based biosensor enhanced with melanin for the early detection of cardiac ischemia by monitoring hypoxanthine levels. Hypoxanthine is a key biomarker that increases during oxygen deprivation in cardiac tissues. The proposed biosensor offers a low-cost, portable, and rapid diagnostic tool suitable for point-of-care applications. By leveraging the conductive and biocompatible properties of melanin, the sensor’s sensitivity and stability are significantly improved.
Creating an Outfit Combination Completion Based Recommendation System
Role: Project Manager | Grant: 7230156
The project focuses on building an intelligent recommendation system that suggests visually and stylistically compatible clothing items to complete an outfit. Unlike traditional recommendation engines that rely on simple similarity or co-purchase patterns, this system analyzes fashion images and outfit compositions to understand visual harmony and contextual pairing. By leveraging deep learning models and computer vision techniques, the system identifies key attributes such as color, texture, shape, and category.
AI-Based Analysis of Microscope Images: Organ-on-Chip
Role: Researcher | Org: InitioCell & ZoiData
This project aims to develop an AI-powered software solution for the automated analysis of microscope images generated by organ-on-chip systems. These platforms replicate key physiological functions of human organs, providing a powerful tool for biomedical research and drug testing. The software being developed is capable of performing automatic image segmentation, cell counting, and statistical analysis. The integration of artificial intelligence ensures objective, scalable, and reproducible analysis.
Automatic Grading of Student Music Exercise Performance
Role: Researcher | Org: Izmir Democracy University
Address the problem of “automatic assessment of student music performances“, for two types of musical exercises: melody repetition/imitation and rhythm repetition/imitation. Our research implemented a system that estimates fundamental frequency series and chroma feature matrices, matching them using a dynamic time warping (DTW) algorithm. For rhythm, we applied metric learning using a Siamese neural network to directly learn from onset positions. The system aims to assess student performance as accurately as a music instructor.