Abdullah Saydemir

Merhaba!

I am interested in machine learning and applications. In ML/AI, I have improved state-space models to make them work under noise, developed generative neural networks for turbulent flow simulations in 3D, created data-driven approach for Lattice-Boltzmann simulations, Laplacians for graph neural networks, graph neural PDEs for protein essentiality prediction, and drug scoring models for cancer treatment. I also worked on non-gradient optimization during my bachelor’s years.

I’m targeting top-tier research opportunities.

Interests
  • Machine Learning
  • Applied ML
  • Natural Sciences
Education

  • MSc in Computer Science, 2025

    Technical University of Munich, DE

  • BSc in Computer Science, 2022

    Özyeğin University, Türkiye

  • Non-Degree Exchange Student, 2021

    Oregon State University, USA

Publications

M. Lienen, A. Saydemir, S. Günnemann (2025). UnHiPPO: Uncertainty-aware Initialization for State Space Models. ICML 2025.

A. Saydemir *-, M. Lienen *-, S. Günnemann (2024). Unfolding Time: Generative Modeling for Turbulent Flows in 4D. AI for Science Workshop @ ICML 2024.

DOI

M. Elyasi, M. E. Simitcioglu, A. Saydemir, A. Ekici, O. Örsan Özener, H. Sözer (2023). Genetic Algorithms and Heuristics Hybridized for Software Architecture Recovery. Automated Software Engineering Journal.

Project DOI

M. Elyasi, M. E. Simitcioglu, A. Saydemir, A. Ekici, H. Sözer (2022). HYGAR: A Hybrid Genetic Algorithm for Software Architecture Recovery. In Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing (SAC ‘22).

Project DOI

A. Saydemir, M. E. Simitcioglu, H. Sözer (2021). On the Use of Evolutionary Coupling for Software Architecture Recovery. 15th National Software Engineering Symposium of Türkiye (UYMS 2021).

Project DOI

Research Experience

 
 
 
 
 
Research Assistant
DAML Group at TUM
Apr 2023 – Oct 2024 Munich
  • Worked on generative diffusion models for turbulent flow simulation with Navier‑Stokes equations in 4D. (10/2023 - 10/2024)
  • Integrated non-conventional Laplacians to spectral GNNs and measured their performance in different heterophily levels. (04/2023 - 10/2023)
  • Surveyed the origins of and sota graph neural PDEs. (04/2023 - 10/2023)
 
 
 
 
 
Research Assistant
Chair of Aerodynamics and Fluid Mechanics at TUM
Sep 2023 – Aug 2024 Munich
  • Modeled collision operators in Lattice‑Boltzmann equations with physics inspired geometric deep learning models as part of interdisciplinary project.
 
 
 
 
 
Research Assistant
Software Engineering Lab at Özyeğin University
Mar 2021 – Jan 2023 Istanbul
  • Built automated systems for investigating the effects of evolutionary coupling in software architecture recovery.
  • Implemented encoder & decoder functions facilitating arithmetic crossover in genetic algorithms for a continuous optimizer.
 
 
 
 
 
Intern
Acıbadem University
Feb 2022 – Mar 2022 Istanbul
  • Worked on drug repurposing for personal prescription for lung adenocarcinoma, breast cancer, and kidney cancer.
  • Coded a drug scoring function using tissue specific GTEx, essentiality, and topological data.
 
 
 
 
 
Teaching Assistant
Özyeğin University
Mar 2019 – Aug 2021 Istanbul

Assisted with preparing course material, conducting tutorials, delivering office hours.

  • Discrete Mathematics for Computer Science (09/2020 – 09/2021)
  • Differential Equations (03/2020 – 06/2020)
  • Calculus II (03/2019 – 06/2020)
  • Calculus I (03/2019 – 06/2020)

Work Experience

 
 
 
 
 
Software Engineer
Datart LLC & Komatsu LTD
Mar 2022 – Mar 2023 Remote
  • Developed the fuel estimation algorithm for haul trucks for Komatsu & Modular Mining.
  • Coded an estimator that produces possible missions and refuel windows in the future using the prior missions of a truck.
 
 
 
 
 
Embedded Software Engineer
QuantWifi AS
Mar 2021 – Jan 2022 Istanbul
  • Coded an audio analyzer in Python using PESQ 862 to find how good an audio file is.
  • Contributed to WiFi analyzer that scans the network and extracts relevant information.
  • Implemented a testbed with prplMesh, MiniNet and other hardware simulators on MIPS architecture.
 
 
 
 
 
Intern, Treasury and Investment Operations
Fibabanka AS
Jul 2019 – Sep 2019 Istanbul
  • Examined and compiled financial information obtained from ledger accounts and bank statements.