I am a gravitational-wave astrophysicist specializing in the search for continuous gravitational waves—faint, long-duration signals from non-axisymmetric spinning neutron stars that have yet to be detected. Their discovery would reveal new insights into the physics of dense matter and neutron star interiors. The core challenge lies in extracting weak signals buried in strong instrumental noise, requiring cutting-edge data analysis and computing techniques.
I am eager to learn from fellow astronomers at this Forum, but I also believe that exchanging ideas with participants from other fields could also help my research. Discussions with mathematicians may inspire new techniques for signal detection in our search. Likewise, the success of machine learning in biomedical applications—especially in handling noisy spectrogram and image data—suggests powerful tools we could learn and adapt to gravitational-wave data analysis. This aligns closely with my current research, which leverages machine learning techniques to detect continuous gravitational waves.
The Forum provides an excellent environment for such idea exchange. I would be honored to contribute to and learn from this vibrant community, and as a scientist from Hong Kong, I would be especially happy to help advance the development of Hong Kong’s academic landscape.

University of Michigan

My current research focuses on enhancing the robustness and sensitivity of searches for continuous gravitational waves—faint, persistent signals emitted by non-axisymmetric spinning neutron stars. Young neutron stars are promising sources for these waves, but they often experience multiple glitches annually. These glitches introduce abrupt frequency jumps in the signal, increasing the likelihood that coherent template-based searches will miss them. To address this, I am developing a robust pipeline that accounts for glitches by analyzing how the statistical properties of detection statistics vary with the number of glitches, ensuring these signals are not missed.
Additionally, I am leveraging machine learning techniques to tackle the all-sky search, which is considered the most computationally intensive. Specifically, I am building a U-Net model to denoise noisy spectrogram data, uncovering signals hidden by the noise. This approach aims to accelerate the all-sky search while using reasonable computational resources. The denoised spectrograms could also enable rapid parameter estimation from the data, enhance the efficiency of signal validation, and improve the inferred model parameter uncertainty in the follow-up analysis.