Our Technology

Ultra-Low LatencyHuman-LevelVoice AI

652M+ voice data and sub-second ultra-low latency
for real-time conversation in education, public services, contact centers, senior care, and more.

Core Technology

Differentiated Core Technologies

Four core technologies for trustworthy, universally accessible AI

<1s

Ultra-Low Latency

Sub-second ultra-low latency real-time conversation so users never feel delay.

Average response under 1 second
Natural two-way conversation
Streaming-based real-time processing
PSTN

PSTN Integration

Zero-Barrier Access
No internet required. We connect the existing telephone network (PSTN) directly to AI for service without device barriers.

No App — No app install or sign-up
Any Device — Use with a regular phone right away
RAG

Hallucination-Free Architecture

For fact-based accurate responses,
we reference external knowledge and verified data
in real time — a core technology for trust.

STT/TTS

Speech Recognition & Synthesis

Best-in-class STT in unstructured noise and natural TTS for human-like voice.

Robust in diverse noise environments
Natural voice synthesis
Real-time processing optimization
Accuracy & Trust

Trustworthy AI

We deliver only accurate, reliable information — what people, enterprises, and public institutions care about most.
Technology designed to remove adoption risk and support confident decision-making.

RAG (Retrieval-Augmented Generation)

Our RAG system addresses hallucination in generative AI and is designed so that AI delivers accurate, trustworthy information that enterprises and public institutions value most.

Real-time database lookupSearch internal manuals, FAQs, and up-to-date sources in real time to generate accurate answers
Hallucination source blockingEvidence-based answers to eliminate the risk of incorrect information
Domain specializationAccurate, reliable answers in tax, accounting, legal, and other specialized fields

Multi-LLM Support

Connect multiple state-of-the-art language models in parallel and choose the best AI model for each use case.

Multiple LLM integrationSupport for latest models including GPT-5, Gemini, EXAONE
Purpose-based optimizationSelect models by use case: education, counseling, assistant, and more
Stable integrationReliable connection between PSTN and LLMs
Data Assets

Irreplaceable
Voice Data Assets

652M+

Voice Data Sets

The largest speech data in the market.
Best-in-class STT that accurately recognizes
even in unstructured noise environments.

We hold the largest speech dataset in Korea (650M+ utterances), enabling accurate voice recognition across diverse environments. We specialize in hard-to-recognize data such as non-native children and silver-generation speech, and this unique asset creates a defensible competitive advantage.

  • ·Unstructured noise robustness — Best-in-class STT in everyday noise
  • ·Diverse age and dialect data — Full coverage of ages and regional dialects for all users
  • ·Continuous learning from live service — Ongoing improvement from 170K+ real users
  • ·Globally validated data — Voice data validated in Korea, Japan, and Vietnam
Our Team

Verified technical capability and
professional experience building AI services

This capability is implemented as a PSTN-integrated network and high-reliability AI architecture, running stably in production.

Core Development Team

Three core full-stack developers and five mid-level developers with Kakao VoiceTalk experience, covering frontend, DevOps, service planning, backend, and UX design as one team.

KakaoTalk FaceTalk & VoiceTalk development
WebRTC-based video solution operations
Generative AI technology and hands-on experience

Technical Capabilities

Voice AI experts with 652M+ voice data, STT/TTS optimization, Hands-off AI, and RAG and Multi-LLM technology.

Real-time voice AI system design and deployment
Large-scale PSTN-based service operations
Trustworthy generative AI deployment
Stable production operations and continuous improvement
Architecture

Architecture Built by Experts

A team with Kakao VoiceTalk experience designed
large-scale traffic distribution and zero-downtime operations.

RAG-based architecture blocks hallucination at the system level and keeps AI reliable in production.

PSTN-Integrated Network

PSTN-integrated infrastructure delivers AI services stably even without internet; large-scale traffic distribution and redundant design support 365-day zero-downtime operations. We built an inclusive AI environment that works with 2G phones and landlines.

PSTN gateway and call-handling infrastructure
Large-scale traffic distribution
Regionally optimized edge servers
Redundancy and failover for zero-downtime

Zero-Downtime & High Availability

High-availability architecture ensures 99.9%+ uptime with automatic recovery from failures. RAG-based retrieval and evidence-joining are integrated at the architecture level to block hallucination in production.

Automatic failure detection and recovery
Load balancing and scaling
Real-time monitoring
RAG-based hallucination blocking (architecture level)