Tolgay Atınç Uzun

Machine Learning Scientist

Machine learning scientist working on LLM-driven AutoML and neural architecture search.

About

I am a machine learning scientist at the Computer Vision Lab, CAIDAS & IFI, University of Würzburg. My research focuses on LLM-driven AutoML: using large language models to design, generate, and optimize neural network architectures in a closed loop.

I graduated as the 2nd ranked student of TED University's Faculty of Engineering (GPA 3.94/4.00) with a background in simulation software (C++, Unreal Engine), computer vision, and image processing.

Germany

Skills

Experience

Machine Learning Researcher

Computer Vision Lab, CAIDAS & IFI, University of Würzburg · 2025 - Present
  • First author of two papers on closed-loop LLM-driven channel configuration search for vision models (arXiv 2601.08517, 2607.20516).
  • Contributor to NNGPT, an LLM-powered AutoML framework that generates, evaluates, and learns from neural network architectures.
  • Contributor to LEMUR 2, a dataset of 14,000+ architectures and 750,000+ training records for LLM-driven AutoML.

Software Engineer

SimBT Simülasyon Bilim ve Teknolojileri · Oct 2022 - Jun 2023
  • Developed simulation software using C++ and Unreal Engine.
  • Worked part-time while completing my bachelor's degree.

Computer Engineering Lab Tutor

TED University · Sep 2022 - May 2023
  • Taught Java in computer engineering lab sessions.
  • Supported students in programming assignments and coursework.

Research and Development Intern

SIMOVATE · Aug 2022 - Sep 2022
  • Worked on image processing and computer vision with OpenCV and CNNs.
  • Developed prototypes with Unity.

Software Developer (Intern)

Monad Yazılım ve Danışmanlık · Jun 2022 - Jul 2022
  • Built web applications with Java and Vaadin.

Software Developer (Intern)

BİTES Defence & Aerospace Technologies · Jun 2021 - Aug 2021
  • Developed desktop applications using C# and Windows Presentation Foundation (WPF).

Projects

NNGPT: Rethinking AutoML with LLMs

Open-source framework turning an LLM into a self-improving AutoML engine for computer vision: architecture synthesis, HPO, accuracy prediction, and NN-RAG.

  • LLM
  • AutoML
  • Computer Vision
  • PyTorch

Closed-Loop LLM Channel Configuration

First-author research: using LLMs to search channel configurations of vision models via conditional code generation with accuracy feedback (CIFAR-100).

  • LLM
  • NAS
  • Research

Education

BSc in Computer Engineering

TED University · 2019 - 2023
GPA 3.94/4.00 · 2nd ranked student in the Faculty of Engineering

Secondary Field in Data Science

TED University · 2021 - 2023

Honors & Awards

2nd Ranked Student in Faculty of Engineering

TED University · 2023

GBYF Second Place Prize in Defense, Cyber Security, Technology and Industry Category

TBD Informatics Association of Turkey · 2023

High Honour Student

TED University · 2020 - 2022

Academic Service

Reviewer · RRPR — Reproducible Research in Pattern Recognition Workshop (ICPR)

Languages

Turkish · Native
English · Full professional proficiency
German · Elementary proficiency

Let's work together

Have a project in mind or want to get in touch?

t.atincuzun@hotmail.com