Supporting a team of 20+ Engineers working on Instagram Relevance Integrity My team works on protecting 100% Instagram users from harmful and unwanted experiences on Instagram largest surfaces - Instagram Connected Surfaces (Feed, Stories) - Instagram Recommendation Surfaces (Reels, Explore, Feed Recommendations) - Threads Trust & Safety, Maching Learning, Recommendation System, LLM

Marketplace Integrity - Trust and safety on Facebook Marketplace - Detection and protection for scam, spam, fraud.

First data scientist hired, build everything from scratch. Product I was responsible for was: -- Jazz Crowd. Jazz Crowd is the premier big data solution for the recruiting and performance management industries. Jazz Crowd will help companies improve processes and results related to hiring peak-performing employees in the human capital management industry. Some of the problems I tackled were: -- Customer Conversion Analysis -- Customer Churn Analysis -- Resume and Candidate Analysis -- Large-scale Job Title Classification -- Large-scale Skills Extraction and Parsing -- Academic Institution Name Entity Normalization Some of the models I used were: Tree-based Model (Decision Tree, Random Forests, Gradient Boosting), Support Vector Machine, Logistic Regression, KNN, Latent Dirichlet Allocation, Naive Bayes, K-Means Clustering, TF-IDF, Vector space model, Topic modeling. Some of the techniques I used were: Python (Numpy, Scipy, Scikit-learn, Matplotlib, Django), Java, R, JavaScript, D3.js, Highcharts.js, MySQL, PHP These U.S. cities are the best bets for new grads seeking jobs Where does your company fall on the time-to-hire spectrum?

As a statistician in the analytics group at Equifax, my major research project was on building a suit of Firmographics inference models using machine learning approaches under the supervision of Dr. Yin. These models had been tested and validated on 500M+ records. My major accomplishments were: • Researched information retrieval and text mining techniques and developed a text feature extraction algorithm for messy textual data. • Built an ensemble classification system of multiple Machine Learning techniques, including Logistic Regression, Support Vector Machine, Naive Bayes, KNN, Random Forests, Gradient Goosting. • Achieved accuracy rate more than 87% on first digit SIC code prediction and 79% on first two digits SIC code prediction. Tools and Techniques used: Python, SAS, R, Bag-of-words, TF-IDF, Ensemble Learning, Logistic Regression, Support Vector Machine, Naive Bayes, KNN, Random Forests, Gradient Goosting, Cross-validation.

• Conducted research on online dimensionality reduction. • Proposed a new algorithm, online sufficient dimension reduction. • Analyzed various high dimensional time series data set • Coded a Python module for real-time multi-gigabyte seismic sensor data processing, cleaning and parsing. • Supervised by Prof. Yao Xie, School of Industrial & Systems Engineering (ISyE)