diff --git a/README.md b/README.md
index 9cfb4c1..7e5661b 100644
--- a/README.md
+++ b/README.md
@@ -25,3 +25,32 @@ In the field of machine learning, such discussions are considered very important
 5. **Seungjun Lee**, and Woojae Seong, Meta-Learned Hamiltonian, Neural Information Processing Systems Workshop on Machine Learning and Physical Sciences, 2020.
 6. Hwiyong Choi, Haesang Yang, **Seungjun Lee**, and Woojae Seong, Type/position classification of inter-floor noise in residual buildings with a single microphone via supervised learning, IEEE European Signal Processing Conference, 2020.
 7. Hwiyong Choi, Haesang Yang, **Seungjun Lee**, and Woojae Seong, Classification of Inter-Floor Noise Type/Position Via Convolutional Neural Network-Based Supervised Learning, Applied Science, 2019.
+
+### Projects
+![Inducing Point Operator Transformer](/fig/IPOT.png)
+- Developing efficient architecture for solving physical systems (_2022.08-2024.02_)
+Main content: Developing a flexible and computationally efficient Transformer to learn physical systems handling continuous functions.
+
+- Proposing a Transformer that is flexible in handling arbitrary discretization %formats 
+and scalable to large discretization sizes.
+
+- Handling arbitrary input/output discretization by processing them in the latent space to avoid quadratic complexity.
+
+- Achieving high accuracy and efficient computational costs in several PDE benchmarks and real-world scenarios.
+
+- Developing robust physics-based neural networks for analyzing dynamic systems}}{2021.03 \textendash 2022.08}{}{\textbf{Main content:} Proposing a method using meta-learning algorithms to model governing equations from similar dynamical systems, enabling rapid learning for new systems even with limited measurement data.}
+
+{- Using a meta-learning algorithm to extract governing rules across similar dynamical systems.}
+
+{- Using Neural ODEs and the Hamiltonian equations to continuously model dynamical systems and utilize the energy conservation principle.}
+
+{- Simulating various systems with different physical properties and initial conditions and validating the capability to efficiently learn new systems even with limited data.}
+
+\entrymid[\textbullet]
+{\textbf{Learning-based sound source classification and localization with limited data}}{2019.03 \textendash 2021.02}{}{\textbf{Main content:} Developing a generative model to robustly estimate the position of sound sources, even from unseen locations within a building during the training. }
+
+{- Applying Zero-shot learning techniques to the problem of sound classification and localization, involving simultaneously classifying the type and location of sound sources.}
+
+{- Using conditional GAN to map sound data to their corresponding types and locations of sound sources and to augment data for unseen locations.}
+
+{- Collecting extensive real-world environment sound datasets, including various types and locations to develop a robust sound generative model.}