RESEARCH
Two SCIE papers in Journal of Cleaner Production (2024 · 2025), authored under AI Korea Inc. affiliation. The same XAI and uncertainty-quantification procedures power our consulting, environmental GIS dashboards, and SaaS.
Lee D., Jeon H. · Journal of Cleaner Production · 529:146805
An enhanced explainable AI for algal-bloom forecasting under climate change, using a multi-run CAM approach to extract consistent model rationales.
View on DOILee D., Lee B. · Journal of Cleaner Production · 473:143457
A reliable AI framework that quantifies uncertainty in deep-learning particulate-matter predictions, supporting policy and operational decision making.
View on DOIRESEARCH FOCI
Narrow and deep — four areas. Each connects directly to one of our four business lines.
01 · XAI
A multi-run CAM approach that extracts consistent model rationales. Core method of our 2025 paper.
→ Business lines: Trustworthy AI Design · Environmental GIS Dashboard
02 · UQ
Confidence bounds on deep-learning predictions, supporting policy and operational decisions. Core method of our 2024 paper.
→ Business lines: Trustworthy AI Design · Environmental GIS Dashboard · Homosilicus
03 · Environmental forecasting
Particulate matter, algal blooms, and other environmental indicators forecast at spatial-temporal resolution and surfaced via GIS.
→ Business lines: Environmental GIS Dashboard
04 · Public-health AI
Avian influenza and other zoonotic disease data, modeled and operated as live dashboards.
→ Business lines: Environmental GIS Dashboard · Job Posting Data (shared pipeline)
HOW WE APPLY
Where each published method is applied across our business lines.
UQ · 2024
Particulate-matter uncertainty
Applied to the air-quality module of our environmental GIS dashboard line (Air-Report). Also a core menu item of our trustworthy-AI design consulting.
XAI · 2025
Explainable algal-bloom forecasting
Carried over into our water and marine forecasting dashboards, and into the reporting pipelines we design for decisions that require accountability.
Data pipeline
Auditable training data
The collection, cleaning, and tagging procedures established in our papers are reused as a standard on both the job-posting pipeline (JobAI) and the environmental observation pipeline.
UQ · XAI → SaaS
Homosilicus persona-response reliability
UQ methods provide confidence bounds and multi-run variance checks for persona responses; XAI methods audit the consistency of underlying rationales. Homosilicus business line → · homosilicus.github.io ↗
COLLABORATIVE RESEARCH
Operational performance and data integrity, verified with seven institutional partners.
Primary partners · 3
Environment · climate
Collaboration on the policy relevance of air- and water-quality forecasting.
Public health
Collaboration on modeling and visualizing wildlife disease data, including avian influenza.
Labor · market
Korea Research Institute for Vocational Education and Training — collaboration on the use of labor-market and job-posting data for policy research.
Additional partners · 4
Medical
Samsung Medical Center (SMC)
Regional R&D
BISTEP
Testing · certification
Korea Testing Laboratory (KTL)
Veterinary · industry
Green Vet Inc.
Next step
We deliver trustworthy-AI design grounded in XAI and uncertainty quantification as an engagement. We reply within 1–2 business days.
