Portrait of Riza Velioglu

Riza Velioglu

Ph.D. candidate in machine learning, leveraging expertise in computer vision and generative models to drive real-world innovation in the fashion industry. My research spans object detection, image segmentation, vision-language models (VLMs), latent diffusion models (LDMs) and generative AI (GenAI). Skilled in end-to-end deep learning workflows, model deployment, and cross-modal learning, I’m passionate about translating cutting-edge research into practical applications.


Education

Bielefeld University
PhD in Machine Learning, advised by Barbara Hammer.
May'21 - Present
Bielefeld University (4.00/4.00 GPA)
MSc in Intelligent Systems, advised by Barbara Hammer.
Oct'18 - Apr'21
Frankfurt University of Applied Sciences
(Erasmus) BSc in Electrical Engineering and Communications Technology
Sep'16 - Aug'17
Istanbul Bilgi University (3.59/4.00 GPA)
BSc in Electrical & Electronics Engineering, advised by M. N. Alpaslan Parlakçı.
Sep'13 - Jun'18

Employment

fal.ai (Machine Learning Engineer)
Lightning-fast inference and post-training of generative media models.
Oct'25 - Present
San Francisco, CA, USA
VisionAI GmbH (Co-Founder & Chief Research Officer)
Co-developed a product portfolio adopted by 50+ customers in a year. Secured €3M+ funding and left before VC to focus on Ph.D.
Jan'20 - Aug'23
Bielefeld, Germany
Bielefeld University (Teaching Assistant & Research Assistant)
TA for 'Introduction to': (1) Machine Learning, (2) Neural Networks, and (3) Algorithms of Computer Science.
Feb'19 - Mar'21
Bielefeld, Germany
Migros Ticaret A.Ş. (Engineering Internship)
Developed a predictive analytics tool to forecast daily sales for Migros stores using classic ML methods.
Jun'18 - Jul'18
Istanbul, Turkey
Istanbul Bilgi University (Student Assistant)
Assisted in the preparation of comprehensive lecture materials.
Sep'15 - Jun'18
Istanbul, Turkey
Sunny Electronics (Engineering Internship)
The products produced by R&D have been tested in accordance with the relevant standards.
Jun'15 - Jul'15
Istanbul, Turkey

Selected Publications

MGT: Extending Virtual Try-Off to Multi-Garment Scenarios
R. Velioglu, P. Bevandic, R. Chan, B. Hammer
With a simple yet effective extension to TryOffDiff, we present Multi-Garment TryOffDiff (MGT), a diffusion-based VTOFF model that supports upper-body, lower-body, and dress garments.
ICCVW'25
TryOffDiff: Virtual-Try-Off via High-Fidelity Garment Reconstruction using Diffusion Models
R. Velioglu, P. Bevandic, R. Chan, B. Hammer
Introducing Virtual Try-Off (VTOFF), a task for generating standardized garment images, and TryOffDiff, a model that adapts Stable Diffusion with SigLIP-based conditioning for high fidelity and detail retention.
BMVC'25 & CVPR'25-Demo
FashionFail: Addressing Failure Cases in Fashion Object Detection and Segmentation
R. Velioglu, R. Chan, B. Hammer
We present a novel e-commerce fashion dataset to enhance object detection and segmentation models, introducing a basic data augmentation approach to increase model robustness for industrial use.
IJCNN'24
Detecting hate speech in memes using multimodal deep learning approaches: Prize-winning solution to hateful memes challenge
R. Velioglu, J. Rose
Won 3rd prize in the competition by finetunung VisualBERT and using ensemble learning.
arXiv'20