HDC Labs Joins National Project for an "Independent AI Foundation Model" to Accelerate Residential AX

Implementing "Operational AI" Through Residential and Building Data-Based Validation... Competing on Inference Cost Reduction, Model Lightweighting, and On-Site KPIs
Proof-of-Concept for Residential and Building-Specialized Multimodal AI via Consortium Participation... Expanding to Smart Home, Control, and Facility Management Services

HDC Labs will participate in the national project for an "independent AI foundation model" as part of the Motif Technologies consortium. Led by the Ministry of Science and ICT, this project is intended to promote the domestic AI industry, and selected companies will be recognized as representing South Korea's national artificial intelligence (AI) while also receiving development support.

While Motif Technologies will lead AI model design and open-source dissemination, and Moreh and Crowdworks will be responsible for infrastructure optimization and data construction respectively, HDC Labs will oversee the proof-of-concept for multimodal AI (LLM/VLM) services specialized for residential and building environments. Through this, HDC Labs plans to strengthen the core competitiveness of its commercial AI wall pad service, scheduled for launch in the first half of next year, and to begin in earnest the expansion of its residential AX foundation.

This project goes beyond simple AI model development to comprehensively verify technological originality and compliance with policy and ethical standards, and will evaluate in stages not only benchmark performance but also whether the models can be used in real-world environments at a reasonable cost. In particular, it aims to secure a cost structure that is applicable to residential and building operation environments and to accelerate the transition to commercialization through real-world validation.

Ahead of commercializing its AI wall pad, HDC Labs has been testing various small language models (lightweight LLMs) under conditions similar to actual service environments. In this process, it identified not only model performance but also operational burdens such as the difficulty of fine-tuning and inference costs depending on user queries as major challenges. Paek Jongmin, Head of HDC Labs Technology Research Center, said, "For the AI wall pad to operate continuously in customer environments, it requires comprehensive verification that goes beyond performance to include cost, response speed, system stability, and security," adding, "By participating in this project, we expect to secure the potential to develop and tune derivative models such as lightweight, distilled, and quantized models based on the foundation model, and to design a cost structure suitable for real-world services."

Through this project, HDC Labs plans to apply AI technology to actual operating environments in the residential and building AX (Automation Transformation) domain and verify its effectiveness using quantitative indicators. The proof-of-concept will go beyond simple demo-level tests and will be carried out in a way that numerically confirms improvements in on-site safety and management efficiency. Key tasks under review include: video-based safety monitoring (automatic detection and reporting of hazardous situations); optimization of common-area operations (usage pattern analysis and detection of abnormal facility conditions); and automation of complaints and notices (summarization, classification, and recommendations).

HDC Labs will also move away from accuracy-centric performance evaluation and instead adopt on-site, experience-based operational indicators as its core criteria. It plans to directly target user evaluation items through field-oriented indicators such as reduction in false detection rates for incidents, reduction in administrators' working hours, shorter incident handling times, performance in handling simultaneous connections, system failure rates, and AI processing cost per case (inference cost).

Residential and building spaces combine diverse data such as video, sensors, documents, and civil complaints, and their outcomes in areas like safety, convenience, and energy efficiency can be measured numerically. For this reason, they are considered fields with high potential for AI technology adoption. HDC Labs plans to expand the technologies secured through this project across all of its services, centering on the AI wall pad and extending to its smart home app, control systems, and facility management solutions.

HDC Labs CEO Lee Junhyung said, "This participation is not just about research collaboration; it is a practical project aligned with the commercial launch of our AI agent wall pad in the first half of next year," adding, "Based on the foundation model, we will secure lightweight, high-efficiency models that can be applied directly to products and thereby strengthen our differentiated competitiveness in the residential and building AX market."