I am a Computer Engineer and M.S. Data Science researcher interested in developing reliable, interpretable, and data-driven AI methods for complex real-world problems.
My work spans machine learning, explainable AI, computer vision, healthcare analytics, geospatial data, infrastructure systems, and disaster-related applications. I am particularly interested in understanding when machine learning models fail, how well they generalize to unseen data, how distribution shift and data leakage affect reported performance, and how AI systems can be made more transparent and reliable for scientific and societal applications.
- Trustworthy Machine Learning
- Explainable AI
- Robust Generalization
- Distribution Shift
- Leakage-Aware Model Evaluation
- Healthcare & Clinical AI
- Computer Vision
- Human Motion Analysis
- GeoAI & Geospatial Machine Learning
- Remote Sensing
- Multimodal Learning
- Spatio-Temporal Data Analysis
- AI for Disaster Assessment
- AI for Infrastructure Resilience
- Environmental & Scientific Machine Learning
I am interested in building machine learning systems that remain reliable beyond standard benchmark settings. My current interests include model generalization, data leakage, distribution shift, robustness, interpretable prediction, transparent evaluation, and realistic validation strategies.
I have experience working with anonymized clinical data and machine learning pipelines for healthcare-related prediction problems. I am interested in reliable clinical prediction, interpretable models, decision-support systems, multimodal health data, and the responsible use of AI in healthcare.
My previous work includes computer vision and sensor-based analysis for physiotherapy movement assessment using Kinect depth-sensor data. I am interested in visual understanding, human motion analysis, multimodal perception, and AI systems that interact with real-world human activity.
I am increasingly interested in geospatial machine learning, remote sensing, satellite imagery, environmental data, and spatial decision-making. I am particularly interested in applying these methods to disaster assessment, infrastructure resilience, environmental monitoring, and large-scale scientific problems.
I am motivated by the use of AI to better understand and respond to real-world disasters and infrastructure challenges. My interests include post-disaster damage assessment, geospatial risk modeling, infrastructure deterioration, remote sensing, multimodal data fusion, and robust prediction across geographic regions.
Machine learning and SHAP-based analysis of long-term pavement deterioration risk using the U.S. Long-Term Pavement Performance database.
Methods: Random Forest, XGBoost, Gradient Boosting, SHAP, SQL, geospatial analysis
Topics: Explainable AI, infrastructure ML, climate variability, model interpretation
β‘οΈ Repository: https://github.com/sevvaluluss/ltpp-pavement-deterioration
Machine learning analysis of anonymized clinical data for pre-EMG Carpal Tunnel Syndrome assessment.
Methods: Logistic Regression, Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost, SVM
Topics: Healthcare AI, clinical prediction, leakage-aware validation, model reliability
A nationally funded research project focused on automated physiotherapy movement accuracy assessment using Kinect depth-sensor motion capture.
Methods: Python, OpenCV, body-joint tracking, temporal motion analysis, classification
Topics: Computer Vision, Healthcare AI, Human Motion Analysis
Large-scale ML analysis using LTPP data, with emphasis on leakage-aware validation, realistic generalization, and SHAP-based interpretability.
Topics: Trustworthy ML, Explainable AI, Infrastructure Analytics, Generalization
Research involving spatial and environmental data, remote sensing concepts, geospatial analysis, and large-scale environmental datasets.
Topics: GeoAI, Remote Sensing, Environmental AI, Spatial Data Science
I am currently expanding my work toward projects at the intersection of:
- trustworthy machine learning
- multimodal AI
- geospatial and remote sensing data
- computer vision
- environmental and disaster applications
- reliable AI for healthcare and scientific decision-making
Current project directions include:
- post-earthquake building damage assessment using GeoAI
- satellite-based disaster mapping
- multimodal environmental hazard detection
- geographic generalization of machine learning models
- rare-event detection under limited labeled data
- explainable and robust AI for real-world scientific datasets
Python β’ SQL β’ C/C++ β’ R β’ pandas β’ NumPy β’ SciPy
scikit-learn β’ XGBoost β’ LightGBM β’ CatBoost β’ PyTorch β’ TensorFlow/Keras
SHAP β’ Cross-validation β’ Group-aware validation β’ Leakage-aware evaluation β’ Error analysis β’ Robustness-oriented evaluation
OpenCV β’ Kinect depth sensing β’ Human motion tracking β’ Image processing β’ Deep learning
GeoPandas β’ GIS β’ Remote Sensing β’ Google Earth Engine β’ Spatial analysis β’ Environmental data processing
Experimental design β’ Model comparison β’ Statistical evaluation β’ Reproducibility β’ Scientific visualization β’ Technical writing
- M.S. Data Science β Maryville University
- B.E. Computer Engineering β KarabΓΌk University
My academic background combines computer engineering, machine learning, data science, computer vision, and applied scientific research.
- Interpretable machine learning framework for resilient modulus estimation of unbound granular base layers: Insights from the LTPP database β Transportation Geotechnics
- Machine learning-based prediction of permafrost degradation and its implications on geotechnical infrastructure β AI in Civil Engineering
- Global seismic site characterization of nuclear power plant sites using USGS Vs30 data β Nuclear Engineering and Technology
- Resilience framework for critical infrastructure: Artificial intelligence applications in post-disaster ground damage assessment
- Advanced GeoAI
- Remote Sensing
- PyTorch for Computer Vision
- Multimodal Learning
- Spatio-temporal Machine Learning
- Robust and Generalizable AI
- Scientific Machine Learning Workflows
- LinkedIn: https://www.linkedin.com/in/sevval-ulus/
- Google Scholar https://scholar.google.com/citations?user=CQXkZU0AAAAJ&hl=tr&authuser=1
- Email: sulusmemis1@live.maryville.edu
- GitHub: https://github.com/sevvaluluss