We transform legacy C code into memory-safe Rust and validate the correctness of the generated code. As machines write more of our software, the central question becomes how to trust what they produce. 

레거시 C 코드를 메모리 안전한 Rust로 자동 변환하고, 생성된 코드의 정확성을 검증합니다. AI가 코드를 대신 작성하는 시대에 어떻게 신뢰성 있는 코드를 생성하고, 검증할 것인가를 연구합니다. 

Keywords: LLM, Machine Learning, C-to-Rust, Memory Safety, Software Validation

Partners: Virginia Tech (SSRG), University of Houston (PaSS)

Ongoing: 

• Trustworthy System Software Generation and Validation 

Background:

• Energy-aware Real-time Hypervisor Design 

• Real-time Scheduling for Multiprocessors

• Wait-free, lock-free synchronization

We detect anomalies and predict failures in industrial equipment from vibration, acoustic, and time-series signals, with a focus on label-scarce settings and real-time on-device inference. 

산업 설비의 진동·음향·시계열 신호로부터 이상을 조기에 감지하고 고장을 예측합니다. 정상 데이터만으로 학습해야 하는 환경, 실시간·온디바이스 추론이 필요한 환경을 주로 다룹니다.

Keywords: Anomaly Detection, Predictive Maintenance

Partners: ETRI, 지나테크, 프레스토솔루션, LS메카피온

Ongoing:

• Industrial AI and Anomaly Detection 

Background:

• Anomaly Detection for Roads 

• Water Pollution Detection Using a Flying Drone Computer Vision 

We study models that jointly learn and generate across modalities — vision, language, and motion. 

영상·언어·동작 등 서로 다른 형태의 데이터를 함께 학습하고 생성하는 모델을 연구합니다.

Keywords: Image Generation, Video Generation, In-betweening, Action Recognition, Sign Language Translation and Generation

Partners: National Research Foundation (NRF), ETRI

Ongoing:

• Sign Language Translation and Generation

Background:

• Talking Face Synthesis 

• Hand-gesture Recognition 

• Finger Position-aware Virtual Keyboard Design 

• Domain Adaptation and Face Recognition

• New Text Entry 

• Novel Human-computer Interaction