IA/ML Ingénieur
MSc in AI from Korea University. Published at IJCNN & Neural Networks. Now building production CV/ML pipelines for smart city infrastructure at Gractor.
Publications à la une
CW-BASS: Confidence-Weighted Boundary Aware Learning for Semi-Supervised Semantic Segmentation
Ebenezer Tarubinga, Jenifer Kalafatovich, Seong-Whan Lee
Tackled boundary blur and confirmation bias using confidence-weighted and boundary-focused techniques. Achieved 65.9% mIoU on Cityscapes with only 100 labeled images.
FARCLUSS: Fuzzy Adaptive Rebalancing and Contrastive Uncertainty Learning for Semi-Supervised Semantic Segmentation
Ebenezer Tarubinga, Jenifer Kalafatovich, Seong-Whan Lee
Introduced fuzzy labels and lightweight contrastive learning to improve generalization in semi-supervised semantic segmentation.
Projets et recherche
Explorations en vision par ordinateur, modèles génératifs, apprentissage par renforcement et plus encore
Computer VisionDual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning
Voir le projet →
Computer VisionSemantic-Aware Multi-Label Adversarial Attacks
Voir le projet →
Computer VisionSelf-Training for Semi-Supervised Semantic Segmentation
Voir le projet →
Computer VisionScalable Urban Dynamic Scenes (NeRF)
Voir le projet →
Computer VisionSpeech Emotion Recognition
Voir le projet →
Generative ModelsAutoregressive Text-to-Image Generation (Parti)
Voir le projet →Expérience professionnelle
AI Research Engineer
ActuelGractor
Sept 2025 – present · Seoul, Korea
- •Designed and implemented AIoT solutions for smart city systems
- •Developed CV and ML models for several smart city systems
- •Integrated ML models into production systems in collaboration with cross-functional teams
Machine Learning Research Engineer
Pattern Recognition & Machine Learning Lab (Korea University)
Aug 2023 – Aug 2025 · Seoul, Korea
- •Built and deployed novel segmentation models, outperforming baselines by up to 25% mIoU
- •Collaborated with industry partners on R&D, publishing papers and filing a patent
- •Developed object detection and tracking pipelines for autonomous driving
- •Led research in depth estimation, instance retrieval, and dense/sparse matching
Software & AI Engineer
GliT
Jan 2019 – Jan 2021 · Hybrid
- •Led tech strategy and completed 10+ full-cycle projects
- •Built and deployed AI-powered applications using deep learning frameworks
- •Launched Innovation Hub clubs in schools, engaging 250+ students
- •Secured 10+ school partnerships, increasing tech adoption by 40%
Formation
Korea University
Master of Science, Artificial Intelligence
2023 – 2025 · Supervisé par Dr. Seong-Whan Lee
Global Korea Scholarship (GKS)Compétences et outils
ML / Computer Vision
PyTorch, TensorFlow, OpenCV, CUDA, ONNX, MLflow, Docker, NVIDIA Jetson
Vision Tasks
Semantic segmentation, object detection/tracking, depth estimation, dense matching, scene classification
Software Engineering
React, TypeScript, Node.js, Express, FastAPI, Three.js, Prisma, PostgreSQL
Languages
Python, C++, TypeScript, C#, Java, Bash
Certificats
IBM Applied AI Professional Certificate
IBM
Modern Robotics Specialization
Northwestern University
Foundations of Project Management
Semantic Segmentation with Amazon Sagemaker
Amazon
AWS S3 Basics
Amazon Web Services
ML Pipelines with Azure ML Studio
Microsoft
Neuroscience
Emory University
Game Development using Scratch
MIT
Beyond the Lab
5M+ streams across piano house, Amapiano, and dance-pop. Three albums. Charted in 3 countries.
Collaborer
Recherche et ingénierie
Ouvert aux collaborations de recherche, au conseil et aux opportunités en ingénierie IA/ML.