Maum AI, Qualcomm collaborate on internet-free home IoT

CEO Yoo Tae-joon

Maum AI Inc., a South Korean embodied artificial intelligence (AI) company that focuses on AI that learns via interaction with the physical environment, will unveil the results of its on-device AI collaboration with Qualcomm Technologies Inc. at Consumer Electronics Show (CES) 2025. The event kicks off in Las Vegas in January 2025.

Maum AI, which is participating in the exhibition with its own booth, will showcase its flagship product, Seamless Uninterrupted Dialogue Assistant (SUDA), which is a voice conversation solution based on a large language model (LLM).

SUDA serves as the foundation model for Maum AI’s voice solutions.

“Zero-latency voice conversations are the core technology,” Maum AI Chief Executive Officer Yoo Tae-joon said.

He explained that like conversing with a person, the AI continuously processes and responds at the right moment while the other party speaks. Even if the speaker interrupts with new instructions mid-response, SUDA can adapt and provide immediate answers.

SUDA integrates Qualcomm’s QXS6490 processor and the second-generation Qualcomm RB3 platform, enabling speech-to-text (STT), LLM, and text-to-speech (TTS) functions for seamless voice integration directly on the device - without the need for an internet connection.

The Qualcomm QCS6490 processor is optimized for industrial IoT applications such as smart homes, transportation, warehouse management, POS kiosks, connected healthcare, and logistics. With its collaboration with Qualcomm, Maum AI successfully integrated STT, LLM, and TTS capabilities into the processor.

“We plan to showcase home IoT functionality running on Qualcomm chips at CES,” Yoo said. “Without relying on AI speakers or internet connectivity, users can issue voice commands to control household appliances and other devices.”

He added that since the functionality works offline, no personal data is transmitted to the cloud, eliminating hacking risks and ensuring robust security.

The company has reduced total processing latency on devices to less than 1.5 seconds.

“Traditional AI speakers often fail to fully comprehend voice commands,” Yoo said. “Maum AI improves user experience by enabling the AI to execute commands immediately and accurately.”

Maum AI will demonstrate real-time control of home IoT systems using SUDA at CES. The company has incorporated API-calling capabilities into its LLM To implement agent functionalities for controlling sensors and appliances within a home IoT environment, allowing users to easily control all API-compatible sensors, appliances, and home IoT devices using natural language commands.