About me
I am a PhD candidate in cognitive neuroscience at University of Oregon. I combine machine learning techniques and statistical inference to study the neural mechanism of human memory and predict memory related behaviors. Along my doctoral training, I find my enthusiasm in solving problems using both structured and unstructured data, inferring from both classic and state-of-the-art algorithms, and communicating stories and insights derived from the data in business settings.
Detail information about my projects can be found here.
Interests
- Data science
- Machine learning engineering
Education
PhD in Cognitive Neuroscience, 2017 - Expected 2023
University of Oregon
Data Science Specialization, 2021
University of Oregon
MS in Cognitive Neuroscience, 2017 - 2018
University of Oregon
BS in Psychology, 2013 - 2017
Beijing Normal University
Skills
Programming
- Python
- deep learning (tensorflow, keras, pytorch)
- natural language processing (huggingface transformers, nltk, gensim, spacy)
- machine learning (sklearn)
- multivariate data analysis (numpy, pandas, scipy, statsmodels)
- data viz (matplotlib, seaborn, plotly)
- web scraping
- R multilevel modeling (lmer), machine learning (tidymodels), data viz (tidyverse, shiny)
- I am the author and maintainer of R package {roistats} for fast multiple testing data analysis.
- I accomplished a five-course sequence of Data Science Specialization with R, which covers reproducible data analysis, data visualization, functional programming, and machine learning.
- SQL • AWS • Bash • C • MATLAB • CSS/HTML
- Git • Singularity
Experience
Data scientist intern
Intuit
- Built end‑to‑end pipeline for quickbooks semantic search with linguistic features and customized word2vec embeddings (baseline model; covered top 1 use case).
- Reviewed 20+ papers on spoken language understanding & query intent understanding; Partnered with product team to synthesize labeled data.
- Applied transfer learning with BERT base and customized top layers for joint intent detect and slot filling; covered all use cases.