🎓 Pursuing a Master's Degree in Computer Science at the University of Trento, Italy. 🇮🇹
🔍As part of my MSc thesis, I am developing unsupervised, generator-agnostic algorithms for detecting generated or altered media by extracting high-level semantic features, contributing to an EU-funded multimedia forensics project. 🕵️♂🇪🇺
🧠 Exploring neuroscience, particularly system neuroscience, with curiosity and a desire to exploit multi-modal fusion techniques in AI such as contrastive learning 🔬
🌿 I like to think and try new ideas. & When away from the computer, I enjoy literature📚, ice skating❄️, and spending time outdoors.
Here are some of the exciting projects I’ve been working on:
- Test-Time Domain Adaptation: Reimplementation of "MEMO: Test Time Robustness via Adaptation and Augmentation. Applied to the ImageNet-A dataset to handle domain shifts in image classification.
- Multimodal RAG: Designed a secure and CPU-efficient retrieval-augmented generation system for genetic research.
- DeepFake Detection A non-linear classifier using CLIP features and a non-linear classifier for detecting AI-generated multimedia.
- Local Sequence Alignment: Implemented the Smith-Waterman Algorithm for local sequence alignment in bioinformatics.
- Metagenomic Analysis: Characterized the oral microbiome using uSGB analysis.
- Knowledge Graph Engineering: Developed a knowledge graph to map Trentino Healthcare Facilities for improved accessibility.
- Interactive Model Inspection for Multi-Class Multi-Label Language Models: Using Plotly, t-sne method, and embeddings to detect any issues (will add the code)
Feel free to reach out if you're looking for a research collaborator—I enjoy engaging in discussions and am open to volunteering on ideas that have the potential to make a meaningful impact. 🌱
- Email: [email protected]
- Linkedin: Zehra Korkusuz