Uğurcan Akyüz

Master's in Computer Engineering

2019 - 2023

Erciyes University, Kayseri, Türkiye

Thesis Title: Fetal Brain Tissue Segmentation Using Deep Learning

Context: To segment fetal brain tissues precisely and efficiently, state-of-the-art 3D segmentation models and patch-based training approaches have been researched. A new 3D U-Net model based on dilated convolutions has been developed. GitHub link of the project

Bachelor's in Computer Engineering

02 - 06/2018

Thomas Bata University, Zlin, Czech Republic

Program: Erasmus+ Student Exchange Program Student

Bachelor's in Computer Engineering

2014 - 2019

Erciyes University, Kayseri, Türkiye

Final Project: Mini Self-Driving Car Prototype

Task: Development of perception module for self-driving.

Computer Vision Engineer

08/2022 - Present

ICterra, Ankara, Türkiye

• Developed an explainable computer vision solution for multi-label breast cancer classification.
• Contributed to the collection of over 500,000 mammography images from multiple hospitals and devices.
• Implemented data preprocessing pipelines for heterogeneous imaging sources.
• Designed and deployed sampling strategies to address severe class imbalance.
• Trained advanced deep learning models, including Domain Adversarial Neural Networks (DANN), to improve cross-domain performance.
• Performed data distribution analysis, model debugging, fine-tuning, and clinical validation in collaboration with radiologists.
• Achieved an average PR-AUC of 0.80 across diverse datasets, demonstrating strong generalization.
• Maintained up-to-date knowledge of domain generalization, explainability, and mammography-focused research.
• Prepared technical documentation for research project funding and deliverables.

Computer Vision Engineer

07/2021 - 07/2022

Akgun Technology, Ankara, Türkiye

• Developed a real-time polyp detection solution for endoscopy and colonoscopy streams.
• Collaborated with gastroenterologists to evaluate videos and perform annotation/labeling for training datasets.
• Researched and benchmarked object detection architectures (YOLO, RetinaNet, Faster R-CNN).
• Developed and packaged a deep learning system that reached a score of 0.86 mAP running at 22 FPS on the GPU.
• Contributed to the migration of on-premise machine learning solutions to AWS, leveraging cloud capabilities for enhanced efficiency and performance.

Research and Development Engineer

10/2019 - 06/2021

Spark Calibration Services Inc., Ankara, Türkiye

• Developed and deployed automatic measurement recognition software using deep learning, delivered as a RESTful service integrated with Metrology.NET.
• Built an in-house data processing tool that reduced processing time on large Excel files from 45 minutes to a few minutes, improving operational efficiency.
• Designed and implemented the SparkS scripting language (DSL) for calibrating RF and Microwave devices.