Belin
EN DE
Curriculum Vitae

Ful Belin Korukoglu

Bioinformatician and societal computing researcher based in Germany. I develop AI models for biomedical research and large-scale data systems for studying society. I'm interested in advancing scientific discovery through reliable, autonomous AI systems capable of evidence-driven reasoning and decision-making under uncertainty.

01 / Current Work
Computational Pathology 2026 — Present

Topology-Aware Graph-MIL

For my Master's thesis, I develop and evaluate graph-based multiple instance learning models using pathology foundation models to predict molecular characteristics from histopathology images, focusing on robustness and generalization.

Computational Pathology & Machine Learning, Schüffler Lab Chair of Computational Biology, Saarland University

Results comparing DNA methylation-based models with H&E histopathology approaches for predicting BRAF V600E mutation status in melanoma, with methylation-based feature selection reaching a patient-level AUROC of 0.784.
Predicting BRAF V600E status in melanoma: methylation-based feature selection reached a patient-level AUROC of 0.784, ahead of the evaluated H&E models.
  • Graph Neural Networks
  • Multiple Instance Learning
  • Pathology Foundation Models
  • Robustness
Computational Social Science 2026 — Present

Geospatial Data Infrastructure

At I2SC, I develop scalable geospatial data pipelines using Python, FastAPI, PostgreSQL/PostGIS, and Redis/Celery, supporting research across 300,000+ locations in 50+ regions. By optimizing asynchronous processing and distributed workflows, I reduced per-record processing time by 35%, enabling faster and more efficient large-scale computational social science research.

Interdisciplinary Institute for Societal Computing (I2SC)

Chart of local restaurant occupancy around 1. FC Saarbrücken match weekends, rising from 30.8 per cent on non-game weekends to 48.3 per cent on away-game weekends.
Restaurant occupancy around 1. FC Saarbrücken match weekends: 48.3% on away-game weekends against 30.8% without a game.
  • Distributed Pipelines
  • Geospatial Analytics
  • Asynchronous Workflows
  • Priority Queueing
02 / Publications
  1. 2026 Poster

    VOYAGER: Probing Traveller Experience and Reported Discrimination in Google Maps Reviews of Global Airports

    Wiltzius, D., Korukoglu, F. B., Weber, I., & Cannanure, V. K. · Second author

    EAAMO

  2. 2025 Peer-reviewed

    Mapping Ethical Guidelines for AI in Healthcare: A Global Perspective

    Korukoglu, F. B., et al. · First author

    Studies in Health Technology and Informatics, IOS Press

    PubMed

  3. 2025 Preprint

    Assessing Genomic Reproducibility of Read Alignment Tools

    Korukoglu, F. B., et al.

    bioRxiv

    bioRxiv

  4. 2024 Conference

    Benchmarking machine learning algorithms for sepsis mortality prediction

    Chhugani, K. et al., Korukoglu, F. B.

    RECOMB-seq

  5. 2023 Conference

    DNA Methylation Analysis of Healthy and Cancerous Tissues

    Korukoglu, F. B. et al.

    Bioinformatics & Computational Genomics Conference, Moldova

  6. 2021 Poster

    Protein Profile of Brain Endothelial Cells

    Korukoglu, F. B. et al.

    Sabancı University Program for Undergraduate Research

03 / Research Experience

Medical Imaging Reconstruction

Harvard Medical School

  • Implemented and evaluated deep learning pipelines for MRI reconstruction across multiple imaging datasets and undersampling conditions, automating model inference and reconstruction-quality evaluation.

Sepsis Mortality Prediction

University of Southern California

  • Benchmarked classical and deep learning models for sepsis mortality prediction across 16 independent transcriptomic cohorts, investigating cross-cohort generalization, feature importance, and reproducibility of predictive biological signals.

Global AI Ethics Atlas

Emory University

  • Developed a full-stack healthcare AI governance platform for organizing and comparing international regulatory frameworks and ethical guidelines. This research contributed to my first-author publication, Mapping Ethical Guidelines for AI in Healthcare: A Global Perspective (2025).

Genomic Reproducibility Benchmark

ETH Zurich

  • Investigated the reproducibility of genomic read alignment tools and developed automated bioinformatics workflows using Snakemake for systematic benchmarking and reproducible analysis.
04 / Selected Projects

Geospatial Analytics

Saarbrucken Eats

An interactive dashboard combining SQL, Google Maps APIs and layered maps to explore restaurants around the Saarland Informatics Campus by cuisine, rating and distance.

  • Python
  • Flask
  • SQL
  • Leaflet
Code

Computational Genomics

Scalable DNA-Seq Pipeline

Automated analysis of PacBio single-end sequencing data with BLEND and Snakemake, built for reproducible benchmarking of read alignment workflows.

  • Snakemake
  • BLEND
  • Bash
  • HPC
Code

Medical Imaging

Multimodal Cancer Diagnosis

A multimodal PET-CT deep learning model predicting lymph-node metastasis, reaching 80% accuracy on the evaluated cohort.

  • PyTorch
  • Multimodal Learning
  • Imaging
Code

AI Governance

Global AI Ethics Atlas

A full-stack platform for organising and comparing 109 international healthcare AI regulatory frameworks and ethical guidelines, developed at Emory University.

  • Next.js
  • Prisma
  • PostgreSQL
Paper
05 / Technical Toolkit

Programming Languages

  • Python
  • SQL
  • C++
  • R
  • Bash

Machine Learning & Scientific Computing

  • PyTorch
  • PyTorch Geometric
  • TensorFlow
  • scikit-learn
  • NumPy
  • pandas
  • SHAP

Data Engineering & Backend

  • FastAPI
  • Flask
  • PostgreSQL / PostGIS
  • SQLite
  • Redis
  • Celery

Infrastructure & MLOps

  • Docker
  • SLURM
  • GPU / HPC clusters
  • Google Cloud
  • Git
06 / AI Safety & Community

Reliability is a community problem.

Alongside my technical research, I'm interested in the reliability, transparency and societal implications of increasingly capable AI systems — the same questions that make reproducibility and careful evaluation central to how I work.

I previously led Marketing and Outreach at AI Safety Saarland (AISS), a student-led initiative promoting safe, fair and beneficial AI development, where I built the brand identity and coordinated interdisciplinary programmes and a mentorship network at Saarland University.

07 / Contact

Let's connect.

I'm always glad to hear about research collaborations, academic discussions and interdisciplinary projects — in computational pathology, scientific AI, or anywhere the two meet.

fulbelin.korukoglu@gmail.com