Eunhyeok Park

Assistant Professor @ POSTECH

Deep learning-based algorithms show excellent performance in various fields such as image processing, natural language recognition, and recommendation system. However, their utilization is limited due to realistic constraints, e.g., computation cost, power restriction, and delayed latency. Our goal is to maximize the usage of deep learning-based algorithms by easing the constraints through HW/SW co-design and optimization. We focus on developing advanced optimization algorithms, e.g., quantization, pruning, and architecture search, to eliminate the redundancy of deep neural networks. In addition, we try to design a specialized hardware accelerator to maximize the benefit of the optimization methods. Eventually, our team aims to automate system optimization so that everyone can enjoy the benefits of ai applications.

Research Interest

Deep Learning Optimization

Sub-4-bit Quantization
Exploiting Temporal-correlation for Video

AI Acceleration System

Energy-efficient NPU design
Large-scale Heterogeneous AI Acceleration

AutoML for Optimization

Neural Architecture Search for Efficient ML
Automated SW/HW Co-design


23, July, 2021: H. Jung, E. Park, S. Yoo, "Fine-grained Semantics-aware Representation Enhancement for Self-supervised Monocular Depth Estimation" is accepted at ICCV 2021 (October 2021) as an oral presentation !

30, June, 2021: We received a gift from Google related to “Studying quality degradation from quantization.”!


Contact Info


Phone: 082-10-9972-2338