From Real-Time Traffic Monitoring to Healthcare: ACRYL Expands into Medical AI
ACRYL Inc. (CEO: Jin Park), a leading AI technology company, announced today the successful demonstration of its AI platform ‘Jonathan’ integrated with a domestically developed AI semiconductor (NPU: Neural Processing Unit) for real-time road traffic monitoring services.
The project enabled vehicles equipped with domestic NPUs to analyze and predict road conditions in real time, while continuously enhancing AI models via cloud collaboration. This hybrid architecture allowed rapid on-device inference with ongoing accuracy improvements through cloud-based large-scale models and model refinement.
‘Jonathan’ played a key role in this process as an end-to-end AI development platform that supports federated learning and cooperative inference between edge devices and cloud servers. Leveraging Jonathan’s federated learning framework, ACRYL enabled learning outcomes from in-vehicle NPUs to be securely transmitted to the cloud, where aggregated insights from multiple vehicles were used to train more advanced models and redistribute them back to edge devices.
This cycle realized a continuous learning-based traffic monitoring system that improves over time depending on driving environments. Notably, sensitive data such as video and location information remained on the device, with only model parameters exchanged—achieving both real-time performance and data privacy.
This technical validation was conducted under the project titled “Development of a Large-Scale AI Application Software Framework Based on Federated Inference and Learning Between Data Centers and Edge NPUs,” supported by the Ministry of Science and ICT and the Institute of Information & Communications Technology Planning & Evaluation (IITP). The project aims to secure real-world use cases for domestic AI semiconductors and build the foundation for commercial AI services integrating edge and cloud technologies.
This marks Korea’s first successful demonstration of a real-time federated learning and inference framework using domestic NPUs in actual vehicles. It showcases the potential to scale across domains such as autonomous driving, smart cities, and urban mobility services.
Federated learning is particularly valuable in sensitive domains like healthcare, where data privacy is paramount. Building on this milestone, ACRYL plans to accelerate the application of its federated learning-based AI platform in the healthcare sector. The company is already deploying various healthcare services through its AI platform tailored for clinical environments, including integration with ML/LLMOps and hospital information systems (HIS). The Jonathan-based federated and continual learning framework—demonstrated here with domestic NPUs—is ideally suited for privacy-sensitive medical applications such as medical imaging analysis, patient monitoring, and inter-hospital collaboration.
“This demonstration showcases both the technical competitiveness of domestic NPUs and the practical viability of federated learning-based AI service architectures,” said Jin Park, CEO of ACRYL. “The implications for privacy-centric industries like healthcare are particularly promising, and ACRYL will take a leading role in advancing AI healthcare platforms and expanding industry applications.”