VEXPLOR
TECHNOLOGY

Not plausible AI —AI that takes responsibility.

When AI is wrong on a factory floor, the line stops. So we ground every answer in the ontology, hand over control only as far as it has been verified, and pursue autonomy only in factories that have been standardized.

5 TECH GROUPS · PATENTS REG 1 / PENDING 2 · GEN-1 2025 → GEN-2 2026
01TECH MAPFIVE GROUPS · MODELS SPECIFIED

Five branches of technology,
all pointed at the factory floor.

T-AProcess prediction & judgment AIA family of deep-learning models that predict the next state of equipment and processes. Where defect data is scarce, generative models augment it for training. MODELSPREDICTIVE MAINTENANCE LSTM AE+CNN · QUALITY PREDICTION LSTM-CNN+ATTENTION · PROCESS OPTIMIZATION TRANSFORMER+RL · TIME-SERIES FORECAST · DATA AUGMENTATION TIMEGAN, ETC.
T-BAI agent orchestrationAn agent architecture split into three tiers — the upper tier coordinates goals, the lower tier executes the work. Domain templates transplant it to each site. ARCHCEO → SUPERVISOR → WORKER 3-TIER
FUNCTIONSPERCEPTION · PREDICTION · AUTOMATION · COMMUNICATION · GENERATION (5)
TEMPLATESMANUFACTURING-PROCESS · PROCUREMENT/MATERIALS (2 DOMAIN)
T-CKnowledge & reasoning engineThe layer that keeps answers from straying past their evidence. Retrieval cross-checks across five paths; reasoning asks for causation, not correlation. ONTOLOGYCROSS-SOLUTION RELATIONS 311+ (CROSS 730+)
RETRIEVALHYBRID GRAPHRAG ×5 — GRAPH · VECTOR · SQL · DOC + GOT
REASONINGNEUROSYMBOLIC CAUSAL REASONING (PEARL SCM) · 6-TIER MEMORY
T-DAutonomy governance & operationsDiscipline before autonomy. Performance is inspected as a contractual metric, not a promise. AUTHORITYAAF L0 READ · L1 PROPOSE · L2 APPROVE-THEN-EXECUTE · L3 AUTONOMOUS EXECUTE · L4 EMERGENCY RESPONSE
SAFETYSHACL BLOCKS PHYSICS-VIOLATING COMMANDS PRE-EXECUTION · SHADOW MODE VERIFICATION
OPSMLOPS DRIFT AUTO-RETRAIN · CONTRACTUAL VERIFICATION METHOD (MAPE · F1 SPECIFIED)
T-EStandards & integrationIt integrates equipment data into international-standard formats. Collection through open protocols, meaning through the asset administration shell (AAS). STANDARDAAS (IEC 63278) AI SUBMODEL STANDARDIZATION — LED THE NATIONAL REFERENCE-MODEL CONSORTIUM
PIPELINEOPC-UA · MQTT → KAFKA · EDGE GATEWAY
Deep dive — this is how we build a dark factory (from data collection to production) →
02NO HALLUCINATIONANSWERS BOUND TO ONTOLOGY

Every answer,
grounded in the ontology.

A general chatbot produces plausible sentences, but a factory floor demands evidence, not plausibility. VEXPLOR retrieves across five paths at once to cross-check the evidence, and verifies the reasoning path itself as a graph.

Because an ontology already defines how items, equipment, processes and quality connect, an answer can only come from within those defined relationships.

Abstract knowledge-graph sceneFIG.3 — KNOWLEDGE GRAPH
QUERY RETRIEVE ×5 [ GRAPH · VECTOR · SQL · DOC · GOT ] ONTOLOGY CHECK [ CROSS-CHECK 311+ RELATIONS · DISCARD UNGROUNDED PATHS ] CAUSAL [ PEARL SCM — CAUSATION, NOT CORRELATION ] ANSWER + EVIDENCE FIG.4 — HALLUCINATION-FREE PIPELINE · QUALITY ROOT-CAUSE 3H → 30S · EQUIPMENT IMPACT 2H → 1M · RECALL SCOPE 4H → 1M (PRODUCT BENCHMARK)
03GOVERNED AUTONOMYAAF — AGENT AUTHORITY FRAMEWORK

Control is handed over only as far as it has been verified.

AAF splits the AI's authority into five levels. The factory decides how far to allow, and commands that violate physical constraints are blocked by SHACL rules before execution. It is designed to meet the governance requirements of regulated industries (IATF 16949, FDA, ISO 13485).

L0

Read

Data viewing and status queries only

READ-ONLY
L1

Propose

Submits its judgments as proposals only

PROPOSE · HUMAN DECIDES
L2

Approve then execute

Executes after operator approval

HUMAN-IN-THE-LOOP
L3

Autonomous execution

Executes automatically within a verified scope

BOUNDED AUTONOMY · AUDIT LOG
L4

Emergency response

Immediate intervention on safety events

SAFETY OVERRIDE
COMMAND SHACL CHECK [ PHYSICS-VIOLATING COMMAND → BLOCKED PRE-EXECUTION · ALTERNATIVE PROPOSED ] AAF GATE [ ALLOWED-LEVEL CHECK ] EXECUTE AUDIT LOG FIG.5 — CONTROL PATH · HANDOVER: SHADOW MODE (JUDGE ONLY) → APPROVE-THEN-EXECUTE → AUTONOMOUS · AI→PLC DIRECT CONTROL DEMONSTRATED ON PHYSICAL EQUIPMENT [VERIFIED]
04GENERATIONS & STANDARDSGEN-1 2025 → GEN-2 2026
GEN-1 · 2025–

Process deep learning

The three core AIs — predictive maintenance, quality prediction and process optimization — built and applied as field projects.

SITES: NKID · ALPET (IN PROGRESS)
MODELS: LSTM AE+CNN · LSTM-CNN+ATTENTION · TRANSFORMER+RL
GEN-2 · 2026–

VEXPLOR Logic Studio AI Agent

Expanded into an agent platform on top of the ontology. Six sites across Gen-1 and Gen-2 combined.

SITES: JUSTEM · SUNYOUNG KOREA · SPEEFOX GEOMDAN · RPS (4 SELECTED 2026)
ARCH: 3-TIER AGENT · GRAPHRAG ×5 · DOMAIN TEMPLATES ×2

We led the consortium developing the national manufacturing-data standardization reference model (AAS/IEC 63278), and the product is designed to meet the requirements of regulated industries such as IATF 16949, FDA and ISO 13485. Equipment integration follows open standards like OPC-UA and MQTT.

PAT REGESG SMART-FACTORY AUTONOMOUS-CONTROL AI (10-2904772) PAT PENDINGGRAPH RAG (KP26037) · NEUROSYMBOLIC CPS CONTROL (KP26038) R&D LABCORPORATE R&D CENTER NO. 2026111564 ISOISO 9001 CERTIFIED ESCROWSW ESCROW — KOREA COPYRIGHT COMMISSION, 3-PARTY (~2029.09)
Planned program target figures are not shown · Task-speed figures are a product benchmark · VERIFIED = field demonstration complete

See this technology running on real floors.

20+ build sites, documented by name and technical stack.

View cases Request a Diagnosis