Uzay Gökay

I am a first-year PhD student in the Institute of Computer Science at
The University of Bonn supervised by Prof. Dominik Bach and
Prof. Juergen Gall. My research revolves around the video understanding task, with a particular focus on unsupervised representation learning for temporal action segmentation within videos.

Previously, I worked as a research assistant at Fraunhofer SCAI where I specialized in biomedical text mining and supervised by
Prof. Holger Fröhlich. Specifically, I implemented and explored various active learning strategies within the context of BERT-based large language models.

Skills

Python

5 years of experience with common data mining and machine learning tools

Machine Learning

Experience with common packages including Scikit-learn, TensorFlow, PyTorch, Optuna, MLflow

Natural Language Processing

Research experience with transformer-based architectures such as BERT, as well as common packages including NLTK, SpaCy, TextBlob

Experience

 
 
 
 
 
Fraunhofer Institute for Algorithms and Scientific Computing
Research Assistant
May 2021 – August 2022 Sankt Augustin, Germany

AIOLOS Project (Artificial Intelligence Tools for Outbreak Detection and Response)

The goal of the project is to develop a digital platform to allow for early detection of new respiratory pathogens epidemics, monitor their spread, and inform decision-makers on appropriate countermeasures.

My task was to perform opinion mining on German tweets to monitor public perception of measures taken by the government in the context of COVID-19 nonpharmaceutical interventions (NPIs).

 
 
 
 
 
Fraunhofer Institute for Algorithms and Scientific Computing
Master Thesis Student
July 2021 – May 2022 Sankt Augustin, Germany

Thesis Title : Disease-Symptom Relation Extraction from Biomedical Textual Content

The main purpose of this research was to extract relations between diseases and their corresponding symptoms from PubMed abstracts. To achieve this, we created a novel dataset by using various active learning (AL) methods which is a form of semi-supervised learning. After the construction of the dataset, we fine-tuned BERT based models which are pre-trained on the biomedical text.

 
 
 
 
 
Bayer AG
Research Intern
October 2020 – April 2021 Leverkusen, Germany

Biotechnology with Focus on Laboratory Automation

  • Programmed a pipetting robot to implement enzyme activity assay experiments
  • Setted up half-throughput fluorescence assays for enzyme activity screening
  • Implemented a Python-based data analysis and visualization pipeline

Contact

If you want to reach out to me, you can use the form below.
  • u.goekay@uni-bonn.de
  • Hertz Chair for Artificial Intelligence and Neuroscience
    TRA - Life and Health
    University of Bonn
    Am Propsthof 49, Bonn, Germany