Premium: acquisition and retention

Graph Affinity: drive network engagement by generating affinity signals

Tools: Spark, Scala, Python, Java, Azkaban, Tensorflow, A/B testing, Airflow

• Designed and built the 1st generation analytical web app for real-time insights and visualizations in public relations.

• Conducted text mining, sentiment analysis, and topic classification for media trends.

• Developed ETL pipeline to stream metadata from social media platforms.

• Geofencing Security System(CV): object detection/tracking (RetinaNet, optical flow)

• Credit Risk Forecasting: credit limit and default rate forecasting

• Knowledge Search System(CV + NLP): optical character recognition(OCR), question-answering, machine reading comprehension(MRC), learning to rank(LTR), multi-label classification

• Personalization App(Reinforcement learning): Thompson sampling for the cold-start problem recommendation

• Other POC ML & Eng projects: data augmentation(imputation, web-scraping), automatic reporting and dashboarding, real-time tracking system(IoT in healthcare)

* Algorithms: BERT, logistics regression, GRU, CNN, RNN, stacking, BiDAF, kNN, decision tree

* Packages: numpy, pandas, scipy, opencv, sklearn, skmultilearn, spacy, allennlp, elasticsearch, asyncio

* Industries: manufacturing, finance, retailer, pharmaceutical, nonprofit

* Tech stacks: Python, PostgreSQL, MongoDB, Spark, VBA, Tableau, Plotly(html, dash)

* Tools: Azure, GCP(Kubernetes, BigQuery, DataStudio, Firestore), AWS(S3, EC2, EMR, Sagemaker)

Fall 2017: Stat 2 and 20

Fall 2018: Stat 2, 20, 133, 134 and 135

• Introduction to statistical and critical thinking

• Introduction to Probability and Statistics

• An introduction to computational data analysis

Tech Stack: R (tidyverse, metafor, psych, mediation)

Methods: Mediation analysis, Wilcoxon signed-rank test, Meta-analysis

Domains: Psychology, Social Science, Pharmaceutical and Drug-Medication Research

Project Highlights:

• Resolved multicollinearity issues in mediation analysis to enhance model validity and interpretability.

• Recommended and applied nonparametric Wilcoxon testing for evaluating Compassion Cultivation Training Questionnaire responses, addressing non-normality in the data.

• Supported medical research by conducting meta-analyses, including effect size estimation, heterogeneity testing, and synthesis of results using established R packages.