CV

Education

Ph.D. Candidate in Civil Engineering, University of Illinois Urbana-Champaign, 2021 - Present
Focus: Machine Learning, Computer Vision | GPA: 4.0 / 4.0
Master of Science in Civil Engineering, University of Illinois Urbana-Champaign, 2019 - 2021
GPA: 3.98 / 4.0
Bachelor of Engineering in Harbor, Waterway, and Coastal Engineering, Zhejiang University, 2016 - 2020

Research and Engineering Experience

Graduate Research Assistant / Machine Learning Engineer, University of Illinois Urbana-Champaign, 2020 - Present
- Built and deployed production-facing ML systems for field inspection, multimodal sensing, and 3D analysis workflows.
- Developed agent-based and deployment-oriented AI workflows integrating data processing, model inference, and result interpretation across modular components.
- Designed and evaluated large-scale image, video, and 3D pipelines under real-world field conditions.
- Machine Learning-Based Field Ballast Condition Evaluation, 2025 - Present
  - Built and deployed I-BALLAST Mobile, an iOS application for field ballast inspection with real-time image processing and on-device inference.
  - Built a performance-sensitive inference workflow supporting concurrent local inference and higher-quality remote DINO-SAM inference when needed.
  - Owned deployment workflow design, profiling, and latency optimization for large-image inference pipelines based on RTMDet and SAHI.
- Mud Spot Ballast Segmentation, 2024 - 2025
  - Trained a DETR-based detector and fine-tuned SAM on augmented muddy-ballast imagery, improving segmentation performance by about 5% in both mAP and mAR.
  - Validated the DETR-SAM pipeline on real field data, achieving 4.18% MAE in Fouling Index and 3.45% MAE in Particle Size Distribution.
- Simulation-Driven Dataset Generation, 2023 - 2024
  - Built desktop software using the Blender Python API to automate simulation-driven synthetic image and point-cloud generation.
  - Developed reusable data pipelines to improve model robustness under limited labeled real-world data.
- Automated Machine Vision-Based Ballast Scanning, 2020 - 2023
  - Designed and implemented the Ballast Scanning Vehicle (BSV), integrating cameras, 3D scanning, and GPS positioning for large-scale field evaluation.
  - Developed segmentation pipelines using Swin Transformer and Cascade Mask R-CNN.
  - Deployed inspection software adopted by the Federal Railroad Administration for field evaluations.
- Riprap Aggregate Size and Shape Analysis, 2020 - 2022
  - Developed I-RIPRAP 3D, an interactive software system for deep-learning-based 3D aggregate analysis and physical-scale estimation.
  - Delivered software adopted by the Illinois Department of Transportation.

Technical Skills

Languages: Python, C/C++, Swift, Objective-C++, Java, JavaScript
ML / AI: PyTorch, Core ML, ONNX, Vision Transformers, DETR, SAM, YOLO, Mask R-CNN, GCN
Vision / 3D: OpenCV, Open3D, COLMAP, Blender Python API, structure from motion, point cloud processing, multi-sensor fusion
Systems / Data: Spark, Hadoop, MySQL, MongoDB, SQLite, AWS (S3, Lambda, DynamoDB, SageMaker, Bedrock)
Applications: SwiftUI, PyQt, React.js, Vue.js, REST APIs

Selected Publications

BallastAttN: Occlusion-Robust 3D Railway Ballast Characterization using Data Synthesis and Deep Learning

Ding, K., Fan, Q., & Tutumluer, E. (2026). "BallastAttN: Occlusion-Robust 3D Railway Ballast Characterization using Data Synthesis and Deep Learning." Transportation Research Record.

Image-Based Moisture Content Prediction in Railway Ballast Using Deep Learning

Ding, K., & Tutumluer, E. (2026). "Image-Based Moisture Content Prediction in Railway Ballast Using Deep Learning." Geo-Congress 2026, 396-406.

Condition Evaluation of Mud Spots in Railroad Ballast Using Deep Learning

Ding, K., Qamhia, I. I., Hart, J. M., & Tutumluer, E. (2025). "Condition Evaluation of Mud Spots in Railroad Ballast Using Deep Learning." Journal of Infrastructure Systems, 31(4), 04025028.

Investigating Effect of Ballast Degradation on Track Dynamic Behavior using Coupled Train-Track Discrete Element Method Model

Liu, Z., Ding, K., Tutumluer, E., & Hashash, Y. M. (2025). "Investigating Effect of Ballast Degradation on Track Dynamic Behavior using Coupled Train-Track Discrete Element Method Model." Transportation Research Record, 03611981251349423.

Evaluation of Field Ballast Degradation at Ft. Campbell Army Installation Using Computer Vision and Comparison with Ground Penetrating Radar

Kim, Y., Ding, K., Koh, Y., Kong, T., Tutumluer, E., Beasley, J., ... & Harrell, M. J. (2025). "Evaluation of Field Ballast Degradation at Ft. Campbell Army Installation Using Computer Vision and Comparison with Ground Penetrating Radar." International Conference on Transportation and Development 2025, 842-852.

Deep Learning Image Segmentation to Assess Field Ballast Angularity Index

Ding, K., Qamhia, I. I., Hart, J. M., & Tutumluer, E. (2025). "Deep Learning Image Segmentation to Assess Field Ballast Angularity Index." Geotechnical Frontiers 2025, 73-82.

A Comparative Study of New Ballast and Recycled Ballast Generated from Various Maintenance Activities

Tutumluer, E., Wang, H., Husain, S. F., Kong, T., Kim, Y., Ding, K., ... & Li, D. (2025). "A Comparative Study of New Ballast and Recycled Ballast Generated from Various Maintenance Activities." Transportation Engineering, 100321.

Towards Automated Field Ballast Condition Evaluation: Field Validation of the Ballast Scanning Vehicle Capabilities

Luo, J., Ding, K., Huang, H., Qamhia, I. I., Tutumluer, E., Hart, J. M., ... & Sussmann, T. R. (2024). "Towards automated field ballast condition evaluation: Field validation of the ballast scanning vehicle capabilities." Transportation Geotechnics, 48, 101311.

Deep Learning Approach for Automated Railroad Ballast Condition Evaluation

Luo, J., Ding, K., Qamhia, I. I., Hart, J. M., & Tutumluer, E. (2024). "Deep Learning Approach for Automated Railroad Ballast Condition Evaluation." International Conference on Transportation Geotechnics, 49-57.

I-BALLAST: Computer Vision Solutions for Ballast Degradation Analysis Using Deep-Learning and Data Fusion Methods

Ding, K., Luo, J., Huang, H., Hart, J. M., Qamhia, I. I., Tutumluer, E., ... & Sussmann, T. R. (2024). "I-BALLAST: Computer vision solutions for ballast degradation analysis using deep-learning and data fusion methods." Geodata and AI, 1, 100007.

Augmented Dataset for Vision-Based Analysis of Railroad Ballast via Multi-Dimensional Data Synthesis

Ding, K., Luo, J., Huang, H., Hart, J. M., Qamhia, I. I., & Tutumluer, E. (2024). "Augmented Dataset for Vision-Based Analysis of Railroad Ballast via Multi-Dimensional Data Synthesis." Algorithms, 17(8), 367.

Augmented Dataset for Multidimensional Ballast Segmentation and Evaluation

Ding, K., Luo, J., Huang, H., Hart, J. M., Qamhia, I. I. A., & Tutumluer, E. (2024). "Augmented dataset for multidimensional ballast segmentation and evaluation." IOP Conference Series: Earth and Environmental Science, 1332(1), 012019.

Toward Automated Field Ballast Condition Evaluation: Development of a Ballast Scanning Vehicle

Luo, J., Ding, K., Huang, H., Hart, J. M., Qamhia, I. I., Tutumluer, E., ... & Sussmann, T. R. (2024). "Toward automated field ballast condition evaluation: Development of a ballast scanning vehicle." Transportation Research Record, 2678(3), 24-36.

Toward Automated Field Ballast Condition Evaluation: Algorithm Development Using a Vision Transformer Framework

Luo, J., Huang, H., Ding, K., Qamhia, I. I., Tutumluer, E., Hart, J. M., ... & Sussmann, T. R. (2023). "Toward automated field ballast condition evaluation: algorithm development using a vision transformer framework." Transportation Research Record, 2677(10), 423-437.

Kelin Ding

CV