OneLineAI

• Booth Number :  

• Contact Person : Jaejun LEE

• Phone :  +82-10-3354-6321

• Email : jjlee@onelineai.com

• Website : www.onelineai.com

• Address :  14F, 122, Mapo-daero, Mapo-gu, Seoul, Republic of Korea (04156)

• Main Product: OLA, RYNTA

---

Company Overview

OneLineAI is a Seoul-based company that builds AI systems for work where errors carry operational or regulatory consequences. The company began with OLA, its financial AI platform for organizing market data, documents, policies and internal knowledge for customer-facing services, compliance and risk operations.

Experience from those deployments led to RYNTA, a validation layer for enterprise agents. RYNTA checks what an agent is allowed to do, tests the basis of its conclusion, controls release and keeps the evidence needed to reproduce the result. The same approach is now being adapted for manufacturing and engineering tasks that require clear rules, independent checks and accountable human approval.

---

Main Products

• 

• 

  05

• 

• 

• 

• OLA

  OLA is OneLineAI's financial AI platform. It organizes disclosures, audit reports, internal policies, product terms, market and financial data, service logs and other unstructured documents. Before this material is used by financial data, chatbot, personalized reporting, compliance or risk agents, OLA normalizes schemas, checks sources, and blocks missing or abnormal data. This gives each service a governed, traceable source of information rather than a loose collection of files and feeds.

  Running OLA in financial institutions has also required dealing with real operating constraints: inconsistent data definitions, access rights, model and document version changes, source citations, staff approval and audit requests. Through those deployments, OneLineAI has accumulated financial data mappings, domain rules, workflow patterns, validation methods and records of how agents fail in production.

  
  

• RYNTA

  The internal validation framework built to keep OLA reliable became the foundation for RYNTA, an enterprise Agent Control Plane.

  RYNTA sits between an enterprise AI agent and the systems it can read or change. For each task, it follows four stages: Contract, Judgment, Intervention and Evidence.

  The Contract sets the boundaries before work begins: which data and tools may be used, which rules apply, when approval is required and what evidence must be retained. Judgment is deliberately divided across components. A deterministic engine runs repeatable calculations and policy checks. ARI gathers context and coordinates specialist agents. AVI recalculates key results independently, challenges assumptions and records exceptions. This separation is suited to risk calculations, regulatory reporting, model validation and controlled document review, where the basis for a result must be traceable.

  RYNTA tests robustness rather than accepting a plausible answer. In adjacent-problem or logic-transfer evaluation, the reasoning is applied to a structurally related case to see whether it still holds when inputs, constraints or evidence change. Its game-theoretic design gives generation and validation agents different roles, information and success criteria. Agreement is not proof; unresolved differences must pass a deterministic rule or go to review.

  If a check fails, RYNTA can stop the action, route it through an approved path, request a new result or escalate it. Material decisions remain with authorized staff under Human 4-Eyes approval; RYNTA does not replace the person accountable for the decision. Inputs, calculations, model versions, findings, interventions and approvals are retained for replay, audit and incident review.

  
  

  OLA provides the governed financial data, documents and operating workflows used by these controls. Together, OLA and RYNTA support agent-based analysis and orchestration while keeping calculations reproducible, validation independent and accountability with people.

---