CV

Curriculum Vitae

Education

  • 2022-2024
    M.S in Data Science
    New York University, Center for Data Science
    • Machine Learning Engineer | NYUIT- HSRN Lab
    • Data Scientist | NYUIT- HSRN Lab
    • GPA, 3.7/4.0
    • Core Coursework
      • Data Science, Probability and Statistics for DS, Linear Algebra, Machine Learning, Big Data, Deep learning, Computer Vision, Natural Language Processing, Deep Learning in Medicine
  • 2016-2020
    B.S in Medical Laboratory Science and Biotechnology
    Taipei Medical University, Taiwan
    • GPA, 3.9/4.0
    • Core Coursework
      • Biochemistry, Molecular Biology, Cell Biology, Microbiology, Immunology, Genetics,.., etc.

Experience

  • 2024 - Present
    Data Scientist / Machine Learning Engineer
    Lasa Health
    • Built an ML-based clinical support system for 26 chronic pelvic pain conditions using XGBoost and ensemble models, achieving 0.85 AUROC scores and targeting a 30% reduction in diagnosis time.
    • Built an automation bot using Selenium to extract standardized medical codes from public sources, cutting data extraction time from one month to three days and significantly accelerating project timelines.
    • Preprocessed/analyzed 100 billion EHR records from Mayo Clinic using BigQuery SQL and Python, applying NLP to improve feature extraction, increasing model accuracy by 5%
  • 2024
    Research Data Scientist
    NYU Langone Health
    • Ongoing research on "Improving MRI Image Reconstruction via Generative AI"
  • 2023- 2024
    Data Scientist
    New York University IT- HSRN Lab (New York)
    • Performed data mining, data analysis (EDA), preprocessing, and feature engineering on a 100GB dataset.
    • Developed scripts for the ETL process, engineered unstructured network data through stream processing.
    • Designed/Implemented cost-effective, novel intrusion detection system (AE model) for full data center monitoring, surpassing traditional methods limited to internet traffic.
    • Developed scripts for the ETL process, ingesting unstructured network data through the ML pipeline.
    • Led the team to build/test SOTA anomaly detection algorithms (Tensorflow Autoencoder-based methods) achieving a 73% F1 score in identifying network anomalies related to malicious behavior across the entire NYU research network outperforming traditional methods (firewalls).
    • Implemented MLOps best practices using MLflow, Docker, and Kubernetes.
  • 2023
    Data Science/ Commercial Analytics Intern , 3 months
    Medidata Solutions Inc (Medidata)
    • Leveraged analytical expertise to evaluate sales performance for pharmaceutical companies (customers).
    • Developed a Tree-based Marketing Mix Model (Random Forest) to optimize campaign strategies and quantify the effectiveness of diverse promotion channels, including phone, in-person, and email outreach.
    • Utilized curve fitting models to assess the KPI impact of various promotional channels.
    • Allocate marketing budget across media channels, reducing expenditure on less effective channels.
    • Anticipated a 30% increase in sales to target customers through data-driven promotion channel optimization.
    • Collaborated cross-functionally to decode intricate data and extract comprehensive insights.
    • Experience working with large, complex datasets, including data from multiple sources, to extract insights and drive strategic decision-making.
    • Leveraged Snowflake, PySpark, and AWS for data analysis, processing, and model development
    • Created impactful data visualizations via Tableau to simplify complex information for stakeholders.
  • 2022-2023
    Data Analyst
    Viz for Social Good
    • Create visualized content using Tableau, in order to communicate to key stakeholders to raise awareness for social issues and inspire support for non-profit institutions.
  • 2018-2022
    Data Analyst / Medical Technician
    aetherAI, Taipei, Taiwan
    • Conducted SQL analysis on 1M labeled data, elevating prediction model precision.
    • Pioneered aetherAI's HEMA project and oversaw clinical trial data management and familiar with ICD-10.
    • Collaborated across departments to facilitate clinical validation processes for FDA approval.
  • 2018-2022
    Data Science Research Assistant
    Sleep Center of Taipei Medical University Shuang Ho Hospital, Taipei, Taiwan
    • Undertook the preprocess (data wrangling and data cleaning) of structured and unstructured sleep data from 1602 patients with Python.
    • Performed statistical analysis, such as the Shapiro-Wilk test and the Kruskal Wallis H test using Python and SPSS.
    • Conducted various machine learning approaches (including logistic regression, random forest, K-nearest neighbors, naive Bayes, and support vector machine) to classify multiple subtypes of sleep disorders with over 80% accuracy, using Python and scikit-learn.
    • Optimized internal clinical decisions with the developed predictive model.
    • Published a machine learning article "Differentiation Model for Insomnia Disorder and the Respiratory Arousal Threshold Phenotype in Obstructive Sleep Apnea in the Taiwanese Population Based on Oximetry and Anthropometric Features" in Diagnostics, 2021, as the second author.

Projects

  • 10-K Analyzer with RAG Chatbot
    • Built a financial analysis tool combining SEC 10-K filings with Retrieval-Augmented Generation (RAG), leveraging LLMs to deliver a context-aware chatbot. Integrated NLP techniques to extract key financial metrics.
  • Credit Card Transaction Fraud Detection
    • Conducted EDA and applied undersampling/upsampling to address class imbalance.
    • Trained classifiers (Logistic Regression, KNN, SVM, Decision Tree) using AWS S3 and SageMaker.
    • Achieved a 0.98 ROC score in detecting fraudulent transactions.
  • Chest Radiology Report Generation using Deep Learning
    • Preprocessed large amounts (567GB) of X-ray images and electric health records from MIMIC-CXR.
    • Applied transfer learning withBLIP-2 to automate Chest X-ray report generation.
    • Enhanced report quality, achieving a BLEU-1 score of 0.34 through find-tuning(LoRa) and prompt engineering.
  • Video Prediction using Combined PredNet and Unet
    • Enhance video frame prediction using a multi-model approach with PyTorch.
    • Implemented a combined PredNet and Unet model to harness sequential and spatial learning.
    • Built combined model PredNet and Unet having the results of Jaccard Index 0.2
  • Chatbot for NYU CDS Website using OpenAI's GPT-3
    • Utilized Langchain to scrape unstructured data and store text data in a vector database.
    • Leveraged OpenAI's Language Model (LLM) with RAG to develop a student-centric chatbot, enhancing user experience.
    • Applied optimization techniques (prompt engineerin) to ensure precise and contextually relevant responses
  • EDA & Predictive Modeling for Renewable Energy
    • Performed exploratory data analysis (EDA), statistical analysis, and visualizations of renewable energy consumption data, including hydroelectric, geothermal, solar, and wind power sources.
    • Built regression models (LinearRegression, DecisionTree, RandomForest) using Python with an average of 100% (r-squared value) in predicting solar power generation for effective energy planning.
  • Twitter Volume Timeseries Forecasting using AWS Sagemaker
    • Managed the dataset on AWS S3 and its’ infrastructure for efficient storage and rapid experimentation.
    • Utilized DeepAR using Sagemaker to train an auto-regressive recurrent network mode, obtaining a remarkable 0.2 MAPE score for forecasting compared to the baseline model (seasonal naive estimator).
  • IMDB Movie Sentiment Analysis using NLP
    • Enhanced BERT-based model for sentiment text analysis using Hugging Face's Transformers.
    • Designed text transformations to enhance model performance, achieving an 89% accuracy rate.
  • Music Recommendation syste based on the ListenBrainz Dataset
    • Used Spark (PySpark) to analyze big data, revealing customer behavior and preferences.
    • Implemented Spark's ALS algorithm, outperforming the popularity-based baseline by 0.5MAP.
  • EDAs and Predictive Modeling for Medical Insurance Pricing
    • Conducted exploratory data analysis (EDA) to identify key factors influencing medical insurance costs.
    • Built regularized regression models (Ridge, Lasso, Elastic Net) using Python with an average of 70% (r-squared value) in predicting the medical insurance price.
  • Storage Optimization for Commerce Retailer Data
    • Revamped storage procedures, resulting in a substantial boost in Spark query speeds.
    • Conducted extensive experiments, evaluating optimization strategies for scalability across diverse data sizes (ranging from 588 KB to 5.88 GB).

Skills and Tools

  • Coding
    • Python, SQL, R, SAS, C++
  • Machine Learning and Deep Learning Packages
    • Pandas, Numpy, Scikit-learn, PyTorch, TensorFlow, Keras
    • Hugging Face, NLTK, OpenCV
  • Data Visualization
    • Tableau, PowerBI, Matplotlib, Seaborn, Plotly
  • Tools
    • Cloud Computing- AWS (Sagemaker, S3, Bedrock, EC2, Glue)
    • MLOPs - Docker, Kubernetes, MlFlow, Databricks
    • Big Data- Snowflake, Spark, PySpark, Hadoop, dbt
  • Others
    • Git, Github, GitLab
    • Linux
    • VS Code, Jupyter Notebook

Ongoing Research

  • Anomaly Detection
    • Unsupdervised Machine Learning
    • Deep Learning
    • ML Deployement
    • Distributed Systems
  • Improving MRI Image Reconstruction via Generative AI
    • Processing and analyzing medical images (MRI images)
    • Diffusion-based deep learning models

Other Interests

  • Hobbies: Badminton, Anime, Muay Thai, Museum Hopping, Photography