• I am a Research Scientist & Engineer @ Pixery Labs and currently working on diffusion-based image & character generation and media quality enhancement.
• In September 2022, I graduated from my M.Sc. degree @ Koç University, KUIS AI Center. I was a member of Intelligent User Interfaces lab and worked on deep detection and recognition models on drawings, comic books, cartoons and animations.

Recent advancements in text-to-image generative models, particularly latent diffusion models (LDMs), have demonstrated remarkable capabilities in synthesizing high-quality images from textual prompts. However, achieving identity personalization-ensuring that a model consistently generates subject-specific outputs from limited reference images-remains a fundamental challenge. To address this, we introduce Meta-Low-Rank Adaptation (Meta-LoRA), a novel framework that leverages meta-learning to encode domain-specific priors into LoRA-based identity personalization. Our method introduces a structured three-layer LoRA architecture that separates identity-agnostic knowledge from identity-specific adaptation. In the first stage, the LoRA Meta-Down layers are meta-trained across multiple subjects, learning a shared manifold that captures general identity-related features. In the second stage, only the LoRA-Mid and LoRA-Up layers are optimized to specialize on a given subject, reducing adaptation time while improving identity fidelity.

Drawing is one of the instruments that people use to convey stories and share their thoughts and feelings. Since the amount of labeled data is limited in the drawings domain, I utilize a wide range of style-transfer techniques (11 styles from 4 studies) on the real-life face and body datasets COCO & WIDER FACE. Furthermore, drawings contain enormous stylistic differences. Thus, I transfer a real-life detector model to this target domain and benefit from a self-supervised teacher-student training structure from raw drawing data. I provide an upper bound by training a fully-supervised detector model with a mixture of all the available labeled data. The supervised model achieves state-of-the-art performance

Worked on Context-based Face Generation in Golden Age Comics (US Comics between the 1930s-1950s). The model predicts the masked face by giving consecutive comic book panels to the model with a randomly selected face masked at the last frame.

An annotator application implemented with Tkinter that is capable of bounding box drawing, character recognition labeling, speech bubble-face-character asssociation, and much more.