• Developed receptive-field encoding models to predict extracellular IC single-neuron responses with frequency and amplitude modulation tuning features (MATLAB).
• Implemented MLP, CNN, and CKAN models to predict neural activity from single-trial data, enhancing model robustness over traditional trial-averaged approaches.
• Analyzing acoustic stimuli signals with signal processing techniques including cochlear filterbanks etc.

• Conducted multidimensional quantitative analysis of color representations in SOTA computer vision models and the fMRI responses of human visual cortex (Python).
• Performed Representational Similarity Analysis (RSA) to assess the alignment of computer vision model representations with human perception, evaluating alignment using Pearson’s r and Spearman’s ρ and contextualizing the results with a leave-one-subject-out noise ceiling analysis.

• Analyzed large-scale calcium imaging data from 10,000 neurons in the mouse visual cortex to characterize functionally coherent neuronal ensembles engaged during stimulus events.
• Applied Non-negative Matrix Factorization to reduce high-dimensional neural activity into 10 low-dimensional“neural factors,” identifying functionally coordinated neuronal assemblies that are spatially clustered within the cortical architecture.
• Built an XGBoost model using neural factor activity to predict pupil size dynamics, achieving 90.3% accuracy on test data.

• Developed a framework to convert Artificial Neural Networks (ANNs) to Spiking Neural Networks (SNNs), translating continuous neural signals into sparse, spike-based computation for reduced energy consumption (Python).
• Investigated brain-inspired mechanisms like Surrogate Gradient Learning and Temporal Encoding to enhance the efficiency and accuracy of converted SNNs for potential on-chip applications.

• Developed a framework to convert Artificial Neural Networks (ANNs) to Spiking Neural Networks (SNNs), translating continuous neural signals into sparse, spike-based computation for reduced energy consumption (Python).
• Investigated brain-inspired mechanisms like Surrogate Gradient Learning and Temporal Encoding to enhance the efficiency and accuracy of converted SNNs for potential on-chip applications.

• Developed a framework to convert Artificial Neural Networks (ANNs) to Spiking Neural Networks (SNNs), translating continuous neural signals into sparse, spike-based computation for reduced energy consumption (Python).
• Investigated brain-inspired mechanisms like Surrogate Gradient Learning and Temporal Encoding to enhance the efficiency and accuracy of converted SNNs for potential on-chip applications.

• Designed and evaluated a hierarchical Transformer model to address cross-document structure in multi-document summarization, overcoming the redundancy and lack of global coherence typical of flat-sequence processing.
• Built a two-stage framework, replacing an LSTM ranker with a BERT-based encoder to jointly encode title-paragraph pairs and improve salient paragraph extraction.
• Developed global Transformer layers with multi-head pooling and inter-paragraph attention to enable effective context fusion across documents.
• Achieved a ROUGE-1 score of 11.11 using the BERT encoder, significantly outperforming the LSTM baseline and demonstrating improved content coverage and relevance.

• Evaluated the performance of Backpropagation in CNNs versus Prospective Configuration in Predictive Coding Networks (PCNs) on CIFAR-10 image classification and target alignment tasks.
• Developed and trained CNN and PCN models, achieving 64.96% accuracy with CNNs (BP) on CIFAR-10 and demonstrating lower test error for PCNs under specific conditions.

GPA: 3.95/4.0

Courses included:

• Intro to Computational Neuroscience,Advanced Computational Neuroscience
• Advanced Topics in Cognitive Neuroscience
• Natural Language Processing

Courses included:

• Artificial Intelligence Technology
• Practice for Software Product Construction
• Operating Systems