Ebenezer Tarubinga

I build AI systems that go from research to production. Currently an AI Research Engineer at Gractor, designing and deploying computer vision pipelines for smart city infrastructure. Previously published 2 first-author papers at Korea University's Pattern Recognition & ML Lab (MSc in AI, advised by Prof. Seong-Whan Lee), and before that shipped 10+ full-stack products as a software engineer.

I care about closing the gap between what works in a paper and what works at scale—whether that means training segmentation models with 100 labeled images, deploying real-time inference on edge devices, or building full-stack applications from scratch.

Email  /  CV  /  Scholar  /  LinkedIn  /  GitHub

profile photo
Selected Work
/
CW-BASS
CW-BASS: Confidence-Weighted Boundary Aware Learning for Semi-Supervised Semantic Segmentation
Ebenezer Tarubinga, Jenifer Kalafatovich, Seong-Whan Lee
IJCNN 2025

Confidence-weighted pseudo-labels, dynamic thresholding, and a boundary-aware module for semi-supervised segmentation. 65.9% mIoU on Cityscapes with only 100 labeled images.

FARCLUSS
FARCLUSS: Fuzzy Adaptive Rebalancing and Contrastive Uncertainty Learning for Semi-Supervised Semantic Segmentation
Ebenezer Tarubinga, Jenifer Kalafatovich, Seong-Whan Lee
Neural Networks 2026

Fuzzy pseudo-labeling, entropy-based weighting, adaptive class rebalancing, and prototype-based contrastive regularization. New baselines on Pascal VOC and Cityscapes across multiple label partitions.

AgriLet
AgriLet — AI Crop Disease Detection & Education Platform
React · TypeScript · Express · Prisma · SQLite

Full-stack platform with AI-powered disease identification, severity scoring, 6 structured learning modules (13 lessons), community data contributions, and downloadable agricultural datasets.

Digital Twin
3D Digital Twin Traffic Management System
React · TypeScript · Three.js · FastAPI

Production edge AI system for real-time traffic monitoring with Three.js 3D visualization, predictive analytics, and smart city integration.

Research
Backdoor attacks
Backdoor Attacks on CLIP (BadCLIP)
Adversarial ML · Multimodal Learning

Studied stealthy backdoor triggers in CLIP-style models via dual-embedding alignment in the joint visual-textual representation space.

Multi-label attacks
Semantic-Aware Multi-Label Adversarial Attacks
Adversarial ML · Multi-Label Classification

Targeted perturbations for multi-label classifiers that exploit semantic label co-occurrence dependencies.

Segmentation baselines
Semi-Supervised Segmentation Baselines (UniMatch + ST++)
Semi-Supervised Learning · Semantic Segmentation

Systematic evaluation of weak-to-strong consistency and self-training that served as the foundation for CW-BASS and FARCLUSS.

Depth estimation
Monocular Depth Estimation (Depth Anything V2)
3D Vision · Autonomous Driving

Depth estimation and point cloud generation for autonomous driving with an improved depth-to-pointcloud pipeline.

LLM context
Effective Context Length of LLMs (STRING)
NLP · Transformers

Investigated the gap between declared and effective context length in LLMs, analyzing positional-encoding biases and the STRING shifting method.

Software
eLearn
eLearn — E-Learning Platform
React · TypeScript · Express · Prisma · SQLite

20+ subjects, 70+ lessons, progress tracking with study streaks, learning analytics dashboard, and personalized settings. Full authentication, Zod validation, and responsive UI.

Experience
Gractor

AI Research Engineer
Gractor · Seoul, Korea
Sept 2025 – present

  • Production CV systems for smart city AIoT; real-time inference on edge devices.
  • End-to-end ML pipelines from data collection to deployment with cross-functional teams.
Korea University

ML Research Engineer
Pattern Recognition & ML Lab, Korea University
Aug 2023 – Aug 2025

  • 2 first-author publications; segmentation models outperforming baselines by up to 25% mIoU.
  • Autonomous driving patent filed; built detection, tracking, depth, and dense matching pipelines.
GliTech

Software & AI Engineer
GliTech
Jan 2019 – Jan 2021

  • 10+ shipped products including AI-powered applications serving hundreds of users.
  • Led engineering and STEM education initiatives reaching 250+ students.
Education
Korea University

Korea University
Master of Science, Artificial Intelligence
2023 – 2025
Advisor: Prof. Seong-Whan Lee · Global Korea Scholarship

Skills
ML / CV PyTorch TensorFlow OpenCV CUDA ONNX MLflow Docker NVIDIA Jetson
Software React TypeScript Node.js Express FastAPI Three.js Prisma PostgreSQL
Languages Python C++ TypeScript C# Java Bash
Certificates
IBM Applied AI Professional Certificate
Modern Robotics Specialization – Northwestern
Foundations of Project Management – Google
Semantic Segmentation with SageMaker – Amazon
AWS S3 Basics – Amazon Web Services
ML Pipelines with Azure ML Studio – Microsoft
Neuroscience – Emory University
Game Development using Scratch – MIT

Template based on Jon Barron's website.