Contact us

From research,
to real-world applications.

From joint R&D projects to a service: for each project, what Metakium did and with which technology.

Under way · design R&D joint research, - · DCP pre-study, launched

MOTIE design R&D

Designing spaces that respond, with generative AI and kinetic media.

A project building a solution that gives people in broadcast, hospital and exhibition spaces a Smart Experience Space (SES), using interactive kinetic media combined with generative AI. KBS Art Vision leads it; Metakium, as a joint R&D institution, handles interactive content that responds to spatial data and multi-sensory staging with light, sound and scent.

A kinetic media scene of glowing rods rippling in response
Project
Development of spatial design technology applied with interactive kinetic media과제명(국문): 인터랙티브 Kinetic Media 적용 공간 디자인 기술개발
Program
Ministry of Trade, Industry and Energy (MOTIE) design innovation programme · design industry technology development (design-tech core technology)Managed by KEIT (Korea Evaluation Institute of Industrial Technology) · project no. RS-2025-14383049
Period
- · 3 years 9 months over 4 project years · phase 1 of 2 under way
Consortium
Lead: KBS Art Vision · joint research: Metakium, Sign Telecom, Kwangwoon University Industry-Academic Collaboration Foundation
Participation
Joint R&D institutionJoint R&D institution · checked · Source: signed R&D plan, MOTIE design R&D (project no. RS-2025-14383049)
Final goal
A solution that uses interactive kinetic media with generative AI to give people in broadcast studios, hospitals in Korea, exhibitions and similar spaces a Smart Experience Space (SES)Technology readiness level (TRL) 4 → 8
Settings
Broadcast · hospitals · exhibitions · building lobbies · shops and showrooms

Goals by project year · whole consortium

Year 1 · 2025SES design scenarios and rendering, multimodal spatial experience design, flexible kinetic display module control, AI user-interaction technology
Year 2 · 2026SES simulation model, human-factors content, advanced 3-axis control, multimodal generative AI content components
Year 3 · 2027Integrated SES prototype, a testbed linking kinetic structures with content
Year 4 · 2028Validation with users taking part, a system ready for production

Data → analysis → use

Data
Room data (light, temperature, humidity, noise) · where people are and how they move · social trends, weather and visitor data
Analysis
Automatic AI content cueing · location-based hyper-personalized staging · a conversational content engine where generative AI answers visitors’ questions
Output
Multi-sensory staging that combines kinetic media with light, sound, scent and air

Metakium’s research work

Year 1Interactive staging designed from spatial dataContent scenarios that change with light, temperature, humidity and noise · multi-sensory spatial experience design
Year 1Location-based hyper-personalized contentMessages and staging matched to movement and gestures · interactive content for several people at once
Year 2Automatic AI content cueing · conversational content engineAI trained on social trends, weather and visitor data arranges the content
Years 2-4Light, sound and scent content and its validationEmotional light-space and sound-space content · response latency and user satisfaction of multi-sensory staging

Project-year goals and Metakium’s work follow the signed R&D plan (July 2025 submission).

Deep-Tech Challenge Project · manufacturing AI

Taking part in the development work of a next-generation thermal-process consortium.

Metakium is a joint R&D institution, in the development work, of a next-generation thermal-process consortium in the Ministry of SMEs and Startups’ Deep-Tech Challenge Project (ecosystem-innovation track). The consortium is one of seven pre-study (phase 1) teams launched in .

DCP participation details
Layers of sensor data rising above a glowing metal block
89,975process data points1-second log · 15 runs across 3 thermal processes
4.96–6.49℃temperature error on unseen runs (RMSE)15 per-process cross-validations
90%target accuracy for AI recipe recommendationfinal goal of the four-year project

Source: DCP phase 1 (pre-study) joint R&D institution results report · Oct 2026

Program
MSS Deep-Tech Challenge Project (DCP), ecosystem-innovation track
Project
Maximising productivity in advanced manufacturing and building a materials, parts and equipment ecosystem for low-power, high-performance next-generation thermal processes
Field
Whether to show the field
Stage
Pre-study (phase 1) · launched MSS announcement, Original press release (review only) (opens in a new tab)
Participation
Joint R&D institution · development work · sensor and process data analysis
Scope
Scope and deliverables

Phase 2 (final five teams) has not been announced as of .

Data → analysis → use

Data
Process and sensor data
Analysis
Sensor- and process-data-based analysis and application
Model
A power–temperature model that predicts actual power and the temperature of three sample zones from set power, gas flow and starting temperature alonePhysics-based: heat exchange between zones, infrared and gas losses, and sensor lag
Validation
Predicts runs it has never seen with a temperature error (RMSE) of 4.96–6.49℃Train on four runs, predict the fifth · 15 cross-validations across 3 processes
Platform
PRISM, a sensor data capture and replay frameworkReal and virtual sensors on the same slots · live capture, dataset builder, replay

DCP consortium

Partner institutionsInstitutions in the pre-study team
Joint R&DMetakiumSensor & process data analysis
NIA hyperscale AI ecosystem · lead institution

We built hand and foot 3D data and nail and pedicure design data.

Metakium led “Nail and pedicure data”, a beauty and health project in the National Information Society Agency (NIA) hyperscale AI ecosystem programme. We built hand and foot 3D modelling data and design data for virtual nail and pedicure try-on and design curation services as hyperscale AI training data, now published on AI Hub.

Hand and foot models with nail art, with an AI 3D hand analysis panel and nail colour and shape cards floating above
80,000hand and foot pose images2,000 sets · RGB + depth
20,058nail and pedicure design images374,008 labels
2,0003D mesh modelsbuilt from 2,000 hand and foot pose sets

Source: “Nail and pedicure data” dataset on AI Hub (opens in a new tab) · published June 2026

Program
National Information Society Agency (NIA) hyperscale AI ecosystem programme · project 3 (beauty and health)
Project
Nail and pedicure data
Period
- (6 months) · completed
Consortium
Lead institution: MetakiumPartners: Kwangwoon University Industry-Academic Collaboration Foundation (AI modelling) · Votre Nail Co., Ltd.
Output
Hyperscale AI training data · available on AI Hub (June 2026) (opens in a new tab)
Keywords
Hyperscale AI · LMM · K-beauty · nail art · 3D modelling

Data → analysis → use

Data
Hand and foot pose capture (RGB + depth, 500+ people) · nail and pedicure design images
Processing
3D mesh models and texture maps · nail and pedicure coordinate masking labels · quality inspectionAI training and validation models by Kwangwoon University
Use
AI training data for virtual nail and pedicure try-on and design curation services

Metakium’s role

01Project lead · data quality management
02Capture setup and data collection
03Processed data3D modelling data built with authoring tools, with the process automated
04Data inspectionAll raw and processed data checked against defined metrics
Service · mobility

AI that recommends and calculates.

I'm Dealer is a long-term rental and operating lease comparison platform that Metakium plans, builds and runs itself. Users pick a few conditions, AI recommends cars, and the monthly payment is calculated on the spot from each capital company's rates.

About the service
The I'm Dealer app self-quote list and a car's monthly payment screen
10capital companies, quotes collected automaticallyKB, Shinhan, Hana, NH, Lotte, Woori, JB Woori, BNK, iM, ORIX
105car models in the database1,970 trims · 29,322 options
652automated test filesfeature tests plus 11 end-to-end screen tests

Source: I'm Dealer source code and vehicle database update records · as of Nov 2026

Service
imdealer.co.kr (opens in a new tab) · long-term rental and lease comparison and quotes
Type
Metakium's own service · planning, development and operation
Period
Development since Apr 2026 · in operation
Technology
Next.js · TypeScript · PostgreSQL (Supabase) · Google Gemini

Data → analysis → use

Data
Synced car, trim and option database · quotes collected from 10 capital companies · user conditions (registration type, mileage, fuel, budget)
Analysis
Cars ranked by use, budget, fuel and popularity scores · Google Gemini explains each recommendation in plain sentencesOnly cars that can actually be quoted become candidates
Output
Monthly payment from each finance company's rates · PDF quotes · hand-off to consultationCompare deposit, prepayment, term and mileage options side by side

What we did well

01AI car recommendationFive use styles plus “let AI decide”, scored with budget, fuel and sales popularity
02Instant monthly paymentCalculated from each capital company's rate table, including deposit and prepayment
03Automatic quote collectionA dedicated collector gathers reference quotes from 10 capital companies' quote sites
04Paperless document checksDriver's licence, health insurance and business status verified online through CODEF

If you have a problem worth testing,
let’s draw the structure together.

If you are considering joint research, technology validation or an education program, get in touch.

Contact us