Research
I study efficient AI systems for foundation models and embodied agents, focusing on inference, memory, and computation reuse.
Research areas
-
Efficient AI Systems
- Memory systems, KV-cache reuse, and foundation-model serving.
- Computation reuse and low-latency inference for interactive agents.
-
Robot Foundation Models
- Efficient VLA and multimodal inference in continuous perception–action loops.
- Temporal reuse, asynchronous inference, and stale-state mitigation.
-
Embodied AI & Robot Learning
- Reusable robot skills, executable policies, and feedback-driven adaptation.
- Evaluation in simulation and real-world robotic manipulation.
Projects
-
2024–2029
HiPER: High-Performance Exabyte Storage Systems
Project Manager / Graduate Researcher
Samsung Electronics
- Study storage-scale memory for foundation-model embodied agents, storing and reusing KV caches for efficient, reliable action selection.
-
2025–2029
Accelerated Insight Reasoning via Continual Learning
Project Researcher
Funded by MSIT through IITP
- Develop reusable robot-policy functions and structured neural memory to retain validated experience and reduce repeated computation across embodied tasks.
-
2022–2026
Adaptive Personality for Intelligent Agents
Project Researcher
Funded by MSIT through IITP
- Develop embodied agents that adapt reusable skills to changing tasks and robot configurations through code refactoring and feedback-driven policy repair.
-
Jan. 2024
Global Capstone Design
Project Manager
Sungkyunkwan University & Victoria University of Wellington
- Prototyped ML-based MySQL buffer management; presented the work and collaborated with students at Victoria University of Wellington.
-
2020
Industry-Academic Collaboration Project
Project Manager
SK M&N Service
- Led a Python/TensorFlow project for speech-based video summarization.
Publications
Full titles, author lists, venues, and paper links are available in the publication list.