Junlin Huang

Research on Cosmic Reionization of High-Redshift Galaxies Using JWST Data

• Research on Cosmic Reionization of High-Redshift Galaxies Using JWST Data
• Measured spectral indicators such as [O III]/[O II] to estimate the Lyman continuum (LyC) escape fraction
• Developed an automated data-processing pipeline in Python using Astropy to perform spectral normalization and emission-line fitting
• Gained hands-on experience in large-scale data analysis and modeling

Protein Embedding and Structural Analysis

• Deployed the ProCeSa framework for protein sequence analysis, replacing the original MLP architecture with the Evolutionary Scale Modeling (ESM) protein language model
• Automated data preprocessing and batch inference in Python (PyTorch, DGL), enabling large-scale protein sequence embedding and structural analysis
• Conducted performance benchmarking to compare MLP and ESM outputs
• Strengthened expertise in model deployment, representation learning

Star Cluster Image Classification with CNNs

• Processed over 2,000 star cluster images from the Las Cumbres Observatory (LCO) telescope archive
• Built a PyTorch-based deep learning model to automatically classify globular and open clusters
• Improved validation accuracy to 93.2% by applying data augmentation and transfer learning techniques
• Applied Grad-CAM visualization to interpret model decision-making and highlight key image features

Modeling Superconductivity in Magic-Angle Trilayer Graphene

• Investigated superconducting properties of MATTG under metal-contact-induced effects
• Performed self-consistent calculations in Python to model interfacial charge transfer and built-in displacement fields
• Analyzed variations in electronic structure and superconducting critical temperature across different metal–graphene interfaces
• Research results published on arXiv (2111.04451)

Yau High School Science Award – Physics (National Second Prize)

• Independently designed and built an experimental platform to study superhydrophobic surface dynamics
• Employed high-speed imaging (300 fps) to measure droplet motion under varying tilt angles
• Applied Python and OpenCV to quantitatively analyze droplet trajectories and surface contact angles

• Designed silicon-based micro-probe layouts using AutoCAD
• Studied and analyzed the structural, electrical, and mechanical properties of elastic micro-probes
• Gained hands-on experience in CAD drafting and mechanical design