Tian Tian

Tian Tian

About

Detail

Senior Data Analyst at Facebook
New York, United States
Contact Tian regarding: 
groups
Networking

Timeline

work
Job
school
Education
folder
Project

Résumé

Jobs verified_user 0% verified
  • Facebook
    Senior Data Analyst
    Facebook
    Oct 2021 - Jul 2023 (1 year 10 months)
    - Driving User Growth of Internal Web Product using Analytics Techniques
    Involved in key metrics framework discussion and design.
    - Created detailed analysis plan in understanding user behaviors in different stages of their journey.
    - Conducted A/B testing and A/A testing to test hypotheses of product features design's impact on key metrics performance. Conducted user segmentation and cohort analysis for different experiments.

  • Parsons Corporation
    Engineer-Data Analyst(Traffic)
    Parsons Corporation
    Jul 2020 - Oct 2021 (1 year 4 months)
    - Conducted traffic demand forecast using predictive modelling.
    - Conducted population segmentation to study travelers' behavior separately and provided insights on improving key metrics performance for disadvantaged population.
    - Ran classification model (logistic regression) analysis to forecast travel modes split and proposed strategies to increase public transit ridership.

  • Parsons Corporation
    Associate Engineer-Data Analyst(Traffic)
    Parsons Corporation
    Jul 2018 - Jul 2020 (2 years 1 month)
  • U
    Researcher
    University of Michigan-Ann Arbor
    Jun 2017 - Jun 2018 (1 year 1 month)
    - Developed regression models to study the factors that affect work and transit accessibility for the disadvantaged population in Benton Harbor, MI by collecting data from survey and GPS devices assigned.
    - Proposed recommendations to local government to address the equity issue.

  • University of Michigan Transportation Research Institute
    Researcher
    University of Michigan Transportation Research Institute
    May 2016 - Feb 2017 (10 months)
    - Developed Pedestrian and Bicycle safety models by analyzing historical accident data and helped create an user interactive visualization platform local residences can check the pedestrian and bicycle safety level in their neighborhood.
Education verified_user 0% verified
  • Coursera
    Managing Big Data with MySQL Managing Big Data with MySQL
    Coursera
    Jul 2021
  • Coursera
    Data Visualization and Communication with Tableau
    Coursera
    Jul 2021
  • Coursera
    Mastering Data Analysis in Excel
    Coursera
    Jun 2021
  • Georgia Institute of Technology
    Master's degree, Data Analytics
    Georgia Institute of Technology
    Jan 2020 - Dec 2022 (3 years)
  • Georgia Institute of Technology
    Master's degree, Data Analytics
    Georgia Institute of Technology
    Jan 2020 - Jan 2022 (2 years 1 month)
  • University of Michigan
    Master of urban planning, Urban and regional planning
    University of Michigan
    Jan 2015 - Dec 2017 (3 years)
Projects (professional or personal) verified_user 0% verified
  • E
    E-commerce platform products recommendations system using DeepFM model
    Apr 2023 - Jun 2023 (3 months)
    - I completed it by myself from start to end
    • Used deep learning model DeepFM (Deep Factorization Model) on e-commerce dataset consist of customer features, product features and context features to predict if customer will click a product link.
    • Used sklearn, tensorflow and matplotlib library for data preprocessing, model training and model evaluation, tuned the hyper-parameters and got the best AUC score as 0.69.
  • N
    NLU Project – Developing a Simplified Q&A System Like Apple Siri
    Jan 2023 - Apr 2023 (4 months)
    - I completed this by myself from start to end
    • Implemented a Simplified Q&A System from start to end .
    • The Q&A System consists of two parts, a Customer Intent Classifier (Trained using Text CNN model) and a Slot Labeler (Trained using biLSTM model).
    • The features pulled by Slot Labeler are sent to third-party API to generate response.
    • Tensorflow, Keras, Sklearn etc., were used in developing the framework.
    • The system achieved the function of automatically generate reasonable response after customer input a query.
  • N
    NLP project - Kaggle Toxic Comment Classification Challenge
    Sep 2022 - Dec 2022 (4 months)
    - I completed by myself from start to end
    • Developed a Toxic Comment Classification Framework from start to end using Object-Oriented Programming to detect toxic comments. The framework consists of four modules: Preprocessor, Trainer, Predictor, Calibrator
    • Tested different models to compare classifier performance: Naïve Bayse, Text CNN, Text RNN, biLSTM, Transformer. Tested adding pretrained word embedding GloVe and contextual embeddings BERT.
    • A combination of biLSTM with BERT contextual embeddings and calibrator got the best performance with a AUC/ROC score as 0.9. Tensorflow, Keras, Sklearn are the major tools used in this project.