Candidates About

 

David Liu

Summary:

·         Over 6+ years of experience in Machine Learning, Data mining, Predictive modeling and Visualization with large data sets of Structured and Unstructured data in IT and Banking Domain.

·         Adept and deep understanding of Python3.3 with Numpy, Pandas, Scipy, Scikit-learn, matplotlib and NLTK.

·         Proficient knowledge on SQL and NOSQL databases like MySQL 5.x, MongoDB 3.x, Cassandra3.x and HBase 1.2.x.

·         Experience in Big Data technologies like Hadoop Eco-system, Spark 2.x and MapR Streaming.

·         Hands on experience in implementing LDA, Naive Bayes and skilled in Random Forests, Decision Trees, Linear and Logistic Regression, SVM, Lasso/Ridge Regression. Testing and validation using ROC plot, K-fold cross validation.

·         Worked with machine learning algorithms such as Adaboost, GBDT, XGBoost, Gaussian mixture model, Structural equation model and Kalman filter.

·         Strong skills in Statistics methodologies such as Hypothesis Testing, Correspondence Analysis, Principle Component Analysis, ARIMA, GARCH time series analysis and A/B testing.

·         Proficient in building, publishing customized interactive Reports and Dashboards by using Tableau9.4, D3.js.

·         Good knowledge on Recommender Systems, Natural Language Processing and Data visualization.

·         Skills in performing data parsing, data manipulation and data preparation with methods including describe data contents, compute descriptive statistics of data, regex, split and combine, remap, merge, subset, re-index, melt and reshape.

·         Working experience in Cloud Computing technologies such AWS EC2, Google Cloud Computing.

·         Hands on Large Parallel and Integrated GPU Computation Platform by using PyCuda 1.2, OpenCL R3.

·         Experience in AGILE methodologies, SCRUM process and GIT for Version Control.

·         Expertise in handling multiple tasks with an aggressive approach to meet deadlines and create deliverables in fast-paced environments; comfortable in interacting with business and end users.

 

 

CORE SKILLS:

Programming Languages	Python2.x/3.x (numpy, pandas, nltk, scikit-learn, matplotlib), SQL, JavaScript, R 3.x, SAS 9.x
Statistical Methods	Time Series   ANOVA   Bayes Law   PCA   A/B test
Machine Learning Algorithms	Regression: Linear/Non-Linear, Logistic, SVM, Regression tree Classification: KNN, Naive Bayes, SVM, decision tree, random forest, Boosting Clustering: K-means, Hierarchical clustering Others: Collaborative Filtering, Neural Network, NLP, Deep Learning
Database Technique	MySQL 5.x    SQL-Server 2010+    MongoDB 3.x    Cassandra 3.x   HBase 0.98
Big Data Technique	Hadoop 2.x   Spark 2.x   HDFS 2.x   Hive 2.x   Hbase 1.x  MapR-Streaming
Cloud Platforms/GPU	AWS    Google Cloud    PyCuda 1.2    OpenCL R3
Data Visualization	Tableau 9.4/9.2, D3.js, Python-Matplotlib
Operation System	Mac OS   Windows   Linux(Ubuntu)
IDE’s	PyCharm2017   Spyder 2.1    JupyterNotebook 4.1    Sublime ‎2.0
Other Skills	XML2.x   CSS   HTML 5.2   AngularJS 1.x  Django 1.11

 

 

 

Professional Experience:

 

Uber Inc.  New York City, NY                                                                                               July 2016 - Till Date

Data Scientist

 

Uber Technologies Inc. is a transportation network company headquartered in San Francisco, California, United States, operating in 570 cities worldwide. It develops, markets and operates the Uber car transportation and food delivery mobile apps.

This project aims to optimizing Uber’s dynamic pricing model and analysis user’s behavior by using machine learning and big data analysis methodology. Identify users and uber-partner(drivers) behaviors in different sessions and tag/classify customers by historical data. Use regional fragmentation analysis based upon Geo-location (fixed geographic area) to optimize driver’s distribution on map.

 

Responsibility:

·          Deployed Adaboost, GBDT, XGboost and other machine algorithms to analysis millions customer’s behavior.

·          Parsed data, producing concise conclusions from raw data in a clean, Well-structured and Easily Maintainable format.

·          Used Pandas, Numpy, PyCuda, OpenCL, Scikit-learn, in Python for developing upon Uber’s Parallel and Integrated GPU Computation Platform.

·          Worked on driver’s profiles and historical data to improve both drivers and users experience and developed data-driven approaches to understand user profiles

·          Performed Linear(Nonlinear) and Logistic Regression (SVM, Random Forest) to tag/classify users.

·          Performed K-means clustering and Multivariate analysis in Python and developed Clustering algorithms and KNN that improved Customer segmentation and Market Expansion.

·          Worked on Regional Fragmentation Analysis based upon Geo-location to optimize driver’s distribution on map.

·          Transferred hexagon regional fragmentation analysis to irregular regional fragmentation analysis.

·          Reduced long-term prediction error of key Uber metrics from 35% to 10%, conducted experimentation and optimization on lifetime valuation of Uber users.

·          Designed and manage A/B experimentation and derive business insights from post-hoc analysis.

·          Built high performance MySQL/Hive/MongoDB queries and intuitive dashboards for management, engineering and internal collaborators 

·          Provided data science support to data-driven decision making in product development cycles.

 

Environment: Python3.3, Scikit-learn, PyCuda 1.2, OpenCL 2.1, MySQL5.7, HDFS 2.7, Hive 2.1, Spark 2.1, MongoDB 3.4.

 

 

 

 

TD Bank   New York City, NY                                                                                             Nov 2015 - June 2016

Data Scientist

 

TD Bank is an American national bank chartered and supervised by the federal Office of the Comptroller of the Currency.

 

This project aims to identify and detect credit card fraud by analysis customer’s transactions and consuming habit in real time. Analyzed Credit/Debit card transactions utilizing a fraud detection algorithm and initiated communication with cardholders and Institutions to verify suspect transactions. Maintained a token secured database of fraudulent transaction details and report. Transfer data into human-readable visualization reports.

 

Responsibility:

·         Designed and implemented data-driven debit/credit card fraud risk model with Python and developed fraud risk rules/strategies by SQL Server2016 and achieved Account Takeover Scenario loss reduction by 10% ($3.4MM) per year. 

·         Obtained and transformed principal components features with PCA in the highly-unbalanced dataset and measuring the accuracy using the AUPRC.

·         Real-time Fraud Prediction Using Spark Streaming and batch processing, modularized Spark functions written for the offline machine learning can be re-used for the real-time machine learning.

·         Used MapR-Streams, MapR-DB(HBase API) and MapR-FS

·         Performed Market-Basket Analysis and implemented Decision Trees, Random Forests and K- fold cross validation.

·         Devised Credit Card Fraud Classification system using SVM in Python, TACL and relational database on HP Non-Stop systems to identify risk of payment transactions and classifying normal versus fraudulent transactions improving F-score of the existing system from 0.65 to 0.94. 

·         Responsible for data identification, collection, exploration & cleaning for modeling, participate in model development

·         Effectively prevented fraud activities of large compromise events by ad-hoc analysis of event data and cooperation with Vendors using efficient SQL/Python program in a timely manner. 

·         Models and probability distributions of various business activities either in terms of various parameters or probability distributions, time-series analysis of time-dependent data.

·         Designed rich data visualizations to model data into human-readable from ROC curve, heat map, D3 visualization, Tableau, etc.

·         Performed ARIMA and GARCH time series analysis and Gaussian mixture model.

 

Environment: Python3.2, Hadoop2, Spark1.6, Spark-Streaming, Hbase, HDFS, Hive, Cassandra3.9, D3.js, Matpoltlib, Tableau9.4, SQL Server2016.

 

 

 

 

 

 

Momo Inc.  China                                                                                                                  Mar 2014 - July 2015

Data Scientist

 

Momo (Chinese Tinder) is a free location-based services instant messaging and dating application for smartphones and tablets. The app allows users to chat with nearby friends and strangers.

 

This project aims to increase the matching rate by analyzing user’s preference and behavior. Deploying A/B test for decision making in launching new features. Text mining user’s profile and reviews by NLTK and target in increasing conversion rate and click through rate. Design recommendation system and sentiment analysis for optimize user’s experience.

 

Responsibility:

·         Performed Logistic Regression, Classification, Random Forests and Clustering in Python.

·         Developed the first hybrid recommender containing both content-based and collaborative filter algorithms.

·         Web-scraped over 310,000 reviews and over 19,000 users' ratings using Python, including Request, Beautifulsoup, lxml, CSS/Xpath selector and Anti-Scrapy technology.

·         Built the text processing pipeline containing Tokenization, Lemmatization, TF-IDF, sentiment analysis, Latent Semantic Analysis and Singular Value Decomposition.

·         Hands on and designed Anti-bots and Target-spam system.

·         Utilized MySQL/MongoDB to store user preference and information and deployed the application to Alibaba-Cloud Computing for better performance.

·         Drawing on Experience in all aspects of analytics/data warehousing solutions (Database issues, Data modeling, Data mapping, ETL Development, metadata management, data migration and reporting solutions)

 

Environment: Python2.7, Html5, Css3, JavaScript, Scikit-learn, MongoDB, Cloud Computing

 

 

 

 

Alibaba Inc.  China                                                                                                               Sept 2012 - Feb 2014

Data Analyst

 

Alibaba is a Chinese e-commerce company that provides consumer-to-consumer, business-to-consumer and business-to-business sales services via web portals. It also provides electronic payment services, a shopping search engine and data-centric cloud computing services.

This project aims to analysis customers’ satisfaction and sentiment by 3 billion of product reviews. Designed and Optimized Connections, Data Extracts, Schedules for Background Tasks and Data Refreshes. Co-operate with Bigdata team and marketing team, data visualization for decision.

 

Responsibility:

·         Acquiring data from Taobao (Chinese Ebay) reviews using python web crawler and Sentiment Analysis.

·         Performed text analysis using signals systems to find patterns in customer behaviors along with Weibo (Chinese Twitter)analytics. 

·         Developed the required XML Schema documents and implemented the framework for parsing XML documents. 

·         Sentiment analysis model to classify and predict reviews using NLTK 

·         Created Dashboards(Tableau/PPT) for stakeholders to monitor KPIs

·         Analyzed trends and rankings by linear and multivariable regression to make more effective prediction and product development decisions.

 

Environment: Python2.7, NLTK, Tableau 9.1, PowerPoint, MySQL 4.1

 

                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        

 

 

China Unicom    China                                                                                                                         Feb 2011 - Aug 2012

Business Analyst

 

China Unicom is a Chinese state-owned telecommunications operator in the People's Republic of China and the world's fourth largest mobile service provider by subscriber base. China Unicom mainly engaged in GSM, WCDMA and FDD-LTE standard mobile network services, fixed communications services, domestic and international communications facilities services business.

 

The major responsibility of this position was performing root cause analysis, gap analysis, troubleshooting, problem resolution and change management, as well as possessing of identifying potential solutions and presenting the proposals to the client/management team.

 

Responsibility:

·         Collecting, understanding, and transmitting the business requirements for the project, and translating these into functional specifications along with customization of telecom software products.

·         Gathered business requirements through interviews, surveys and observing from account managers and conducted controlled brain-storming sessions with project focus groups and documented them in the Business requirement document. 

·         Created Use Case Diagrams, Activity Diagrams, and Sequence Diagrams using MS Visio/Excel. 

·         Coordinated with QA team to create the test approach and determine test needs, test environment, test data, resources and limitations.

·         Assisted the QA in performing simple SQL queries for QA testing and data validation.

 

Environment: MS Visio, MS Office(Excel/PowerPoint/Word), SQL-Server 2010

 

 

Education:

Master of Science in Network & Communications Engineer & Services

Bachelor of Science in Telecommunication Engineer