.svg)
OneLineAI
About the Organisation
OneLineAI is a Seoul-based financial AI company building governed AI for banks, securities firms and exchanges. Its OLA AX platform connects internal knowledge, financial data and business systems into one AI workspace: OLA Knowledge Management turns documents into permission-aware, citable knowledge; finance-specific agents cover investment briefings, disclosure and research chat, compliance review, credit and risk analysis and internal automation; and OLA sLLM provides in-house Korean financial reasoning models (4B to 35B) that run on-premise or hybrid. RYNTA, OneLineAI's agent control plane, validates, approves and audits every agent action, and RYNTA RiskOps brings capital, liquidity, credit, climate and AI risk into one controlled workspace where deterministic engines calculate, independent validation challenges and accountable people decide. Clients include Samsung Securities, IBK Securities, Hana Securities, LS Securities, Korea Exchange and Industrial Bank of Korea. OneLineAI also developed KMMLU, HRM8K and HAE-RAE Bench, benchmarks used by major Korean LLM developers.
Products and Services
-
OLA Knowledge Management
Turns scattered internal documents into AI-ready, governed knowledge. Ingests PDF, HWP, PPT, Excel, images and scans with OCR, layout and table parsing, chunking, tagging and embedding. Processes DRM-protected files without removing DRM and deletes plaintext after processing. Hybrid search (keyword, BM25 and vector) with financial-context re-ranking returns answers with source citations and original-text highlights. Includes personal, department and company knowledge scopes, SSO and org-chart-based permissions that sync with HR changes, an admin console, and full audit logs that also record who viewed the audit logs and why.
-
OLA Financial Agents
Ready-to-use AI agents built on real workflows in securities, banking and capital firms. Investment & PB: hyper-personalized market briefings, portfolio issue alerts and PB sales reports. Financial & disclosure chatbot: answers grounded in 20 financial data nodes and DART filings. Research: multi-source analysis of markets, news, filings, reports and financials. Compliance: regulation tracking, suitability and investment-ad review, and mapping from rules to internal controls. Credit & risk: borrower and financial analysis, stress scenarios and review report drafts with independent re-calculation and human approval. Internal automation: reports, meeting notes, translation and controlled actions on internal systems.
-
RYNTA (AI Agent Control Plane)
RYNTA is the control plane that validates and governs financial AI agents from build to run. Agents are defined with approved models, prompts, knowledge and tools, tested in a sandbox for quality, sources, permissions and sensitive data, then approved, versioned and published to a catalog with rollback. At runtime, contracts and policy rules decide which data, tools and actions each agent may use; risky actions are blocked, rerouted or sent for human approval before execution. Every decision is stored as evidence (user, agent version, prompt, model, source, rule, approval and result) for internal audit and regulatory review.
-
RYNTA RiskOps (Controlled AI for Risk Operations)
Built on the RYNTA control and validation layer, RYNTA RiskOps is a risk management workspace for banks that brings capital, liquidity, credit, climate and AI risk into one place. AI agents prepare the work; accountable people decide. Every result follows a controlled path: Data → Calculate (deterministic engine) → Validate (independent AVI validation with criteria, findings and exceptions) → Human review → Evidence, so each figure traces back to its source ledger. The enterprise risk cockpit shows CET1, stress-test troughs, LCR/NSFR, limit breaches and open exceptions on one screen, with 3D views for RWA and capital sensitivity, interest-rate shocks, liquidity gaps by maturity, rating migration, default history and climate exposure by country. AI-risk oversight tracks each agent's activity, permissions, tools and handoffs. Same inputs, same results; runs without external services.
