Innovative AI Projects
Transforming industries with cutting-edge AI in solar energy, energy forecasting, computer vision, robotics, smart agriculture, and smart cities, with video walkthroughs that bring our research to life
Latest Lab Projects
Built by our researchers and made to explore. Watch each project come to life in a short video walkthrough.
PV-Scope
One platform, three scales: multi-resolution solar panel segmentation
Discover, identify and measure photovoltaic systems in satellite, aerial and UAV imagery. An agentic MAPVNet router picks the specialist model trained for each image's scale, so every panel pixel is marked accurately whatever the source.
Highlights
Technologies
SolarMap AI
Find, measure and verify solar installations across Jeju
Maps solar PV systems in satellite imagery of Jeju and compares what the model finds against the official register. Operators can correct outlines by hand, estimate energy production and retrain the model from their corrections.
Highlights
Technologies
PI-Solar Power
Physics-informed solar forecasting you can verify
Asks one question: if a machine-learning forecaster is given the physics of the sun, does it forecast better, and can we prove it? Solar position, clear-sky irradiance and cell temperature set the ceiling, and the same learner is scored with and without physics on unseen data.
Highlights
Technologies
Solar PV Simulator
Configure modules and explore partial-shading scenarios
A step-by-step simulator for photovoltaic experiments: choose a real module or upload your own, set the array layout, irradiance and cell temperature, then explore shading scenarios to see how global maximum power point tracking responds.
Highlights
Technologies
HyRaT-Meta
Forecasting energy demand, renewable generation and curtailment
A research framework that forecasts how the grid will keep supply and demand in balance: energy load, renewable generation, and the curtailment that happens when there is more renewable power than the grid can use.
Highlights
Technologies
ProphetBoost Studio
Forecast electricity demand in one click
Pick a dataset, pick your models, and see how well they predict, with live charts and side-by-side scores. Uses the built-in Jeju Island dataset or your own CSV, and runs several models at once for a fair comparison.
Highlights
Technologies
Learning to Navigate
Goal-directed robot navigation with reinforcement learning
Robots learn to reach a goal while avoiding obstacles, trained with reinforcement learning in the Webots R2025a simulator on both an NAO humanoid and a quadruped platform.
Highlights
Technologies
Smart Farm
Evidence-grounded monitoring and inspection planning
Monitors farm conditions, identifies the areas that need attention and helps plan inspections. The AI shows its evidence, while the operator approves every action, so devices are never changed automatically.
Highlights
Technologies
Our Research Projects
Hybrid Architecture for Real-Time Query-Based Video Instance Segmentation
A real-time vision-language segmentation framework that combines heavy detection on key frames with lightweight mask propagation, achieving near-SOTA accuracy at 190+ FPS.
Technologies
Objectives
LLM for Traffic Digital Twin
Using LLM to automatically generate traffic scenarios for building and optimizing city transportation digital twins.
Technologies
Objectives
Lightweight LLM Technology for Smart Tourism
Developing a low-power LLM-based AI recommendation service for smart tourism with personalized information delivery.
Technologies
Objectives
LLM-Based Smart Farm Service
Creating a lightweight LLM for wasabi smart farming optimization with sovereign AI capabilities and real-time IoT integration.
Technologies
Objectives
Establishing On-Device AI for Drone-Based Solar Panel Inspection
Building a smart preventive maintenance system using on-device AI on drones for automated solar panel inspection and fault prediction.
Technologies
Objectives
AI-Based Roof Edge Extraction for Solar Panel Installation
Combining ViT and LLM technology to identify roof edges and obstacles from aerial imagery for optimal solar panel installation planning.
Technologies
Objectives
Development of Data Management Technology for Sovereign AI BEMS
Using LLM to standardize and normalize diverse BEMS data for optimal building energy management with sovereign AI capabilities.
Technologies
Objectives
Collaborate on Cutting-Edge AI Projects
We welcome partnerships with industry leaders, research institutions, and innovators to push the boundaries of AI technology together.