Yi Liu Email & Phone Number

Los Angeles, California, United States

1. Merged survey and player activity tables and aggregated features to support Match-3 game experience analysis, using Databricks.2. Created dashboards in Qualtrics to monitored KPI vitals, such as player’s NPS, engagement, feedback on product, and game’s operational status.3. Developed a web scraping to automatically get app reviews in SensorTower and import them into Qualtrics using Python. Removed the emoji using ASCII table to avoid column display error.4. Examined the correspondence between player experience and monetization metrics via correlation coefficient using Python, and improved it from 0.5 to 0.7 after adding a 1-week time shift to one metric.

1. Conducted a NER analysis to identify corporation related news from a twitter scandal dataset by distinguishing certain brand names, like “Target”, from their other meanings, achieving accuracy up to 83%.2. Compared the different NER libraries and methods, including Flair, Spacy, NLTK and transfer learning. Transfer learning has the highest accuracy and Flair runs the fastest among these methods.3. Trained deep learning models for classification between scandals and daily news, including BERT, FastText, Transformer and StackedEmbeddings. Tried different test set splits and tuned parameters to improve and compare the performance. The current best model is TransformerDocumentEmbedding, with 85% accuracy.

California, United States

1. Crawled 2TB job posting data from job boards, wrangled them in AWS virtual environment and stored them in SQL database, to support industry analysis2. Established daily refreshed interactive Tableau dashboards explaining industry trend insights for Kern County. 3. Created a job recommendation system for requirement matching between recruiters and job seekers, using content-based algorithms. Optimized it via extracting new features using NLP.

Wuhan

1. Participated in an agronomy project on the correlation between aquatic animal’s body color and its growth speed and body type2. Data extraction: Conducted high-throughout sequencing, to cut off the original gene fragments of each individual, and obtained data of each fragment3. Feature Preparation: Utilized Genome Wide Association Study (GWAS) to find the gene fragments that corresponded to sexual characteristics, and finally spliced into a gene segment that conformed to the target sexual characteristics4. Workflow of analytics: Covering identification, data collection, data cleansing, pattern recognition, result visualization, using R studio.

Doctor Of Philosophy - Phd, Computer Science

As a dedicated PhD student with a focus on Computer & Data Science, I am deeply engaged in exploring the dynamic field of agentic Large.

Master Of Science - Ms, Analytics, 3.8

Bachelor Of Science - Bs, Finance, 3.8