Jingzhi Pang

Carnegie Mellon University

Master of Science (M.S.), Computational Biology

January 1, 2015 – January 1, 2017

Shanghai Jiao Tong University

Bachelor's degree, Biomathematics, Bioinformatics, and Computational Biology

January 1, 2011 – January 1, 2015

Google

Software Engineer

November 1, 2021 – Present

Petuum, Inc.

Software Engineer II (Machine Learning)

February 1, 2018 – October 1, 2021

San Francisco Bay Area

Agile SDE, LLC

Data Scientist

July 1, 2017 – February 1, 2018

San Francisco Bay Area

Carnegie Mellon University

Graduate Research Assistant - SCS - CBD - Langmead Lab

January 1, 2017 – March 1, 2017

Greater Pittsburgh Region

Amyris

Scientific Computing Intern

June 1, 2016 – August 1, 2016

Emeryville, CA

Survival Prediction on Kaggle Titanic

February 1, 2017 – March 1, 2017

·  Set up a Hadoop mapreduce version of KNN classifier from scratch ·  Performed data pre-processing and feature engineering on Titanic dataset, creating 3 new features ·  Improved the accuracy of KNN classifier from baseline of 74% to 90.43%, with precision and recall highly increased ·  Built up Gradient Boosted Trees classifier with xgboost, implemented grid search optimizing parameters in model tuning, improving accuracy from baseline of 76.08 to 88.52%

Active Learning on Image Classification

October 1, 2016 – December 1, 2016

· Trained an active leaner with Query by Committee(QBC) as query strategy and SVM with rbf kernel as classifier, and compare its performance with a baseline learner using random sampling strategy · Achieved meow than 90% accuracy, which is the same as base learner using only 40% of total samples

TSS Recognition Pipeline Construction-"PyCorn"

March 1, 2016 – May 1, 2016

· Participated in the design of "PyCorn"- a genome-wide transcription start sites(TSS) prediction pipeline of Zea mays, as group technical leader · Built up neural network(NN) to predict if the input genome sequence contains a TSS with scikit-learn package · Improved the prediction ability by optimizing the parameters of NN, such as activation function, number of hidden units, etc. · Designed the input module of PyCorn to deal with large input sequence · Achieved 83.9% accuracy on TSS prediction

Neuroscience meets Deep Learning

February 1, 2016 – April 1, 2016

· Performed data pre-processing, transforming functional magnetic resonance imaging (fMRI) data associated with different words of 9 subjects, to three-dimensional(3D) space · Built up three-dimensional convolutional neural network(3D CNN) model, which is composed of convolution layers, max-pooling layers, a dense layer and a logistic regression layer, to predict the neural activation associated with different categories of words · Achieved three times higher accuracy than a random classifier · Compared 3D CNN prediction ability with other basic machine learning models, such as neural network, random forest, etc.

Automated Recognition of Pancreatic Cancer

January 1, 2016 – May 1, 2016

· Built up SVM classification models to determine whether a patient has pancreatic cancer with proteomic data · Enhanced the classification accuracy to above 80% by Implementing feature extraction and false negative control · Implemented unsupervised feature selection with Gaussian Mixture Model(GMM) · Applied "Affinity Propagation", a clustering method, to find the subtypes of pancreatic cancer · Proved statistical correlation between different pancreatic cancer clusters and clinical symptoms with chi-square test

Image Classification of CIFAR-10 dataset

November 1, 2015 – December 1, 2015

· Extracted visual features of CIFAR-10 dataset, which consists of 5000 pictures in 10 classes · Established classification models based on SVM, softmax regression and k-binary logistic regression, and K-nearest neighbors algorithm · Improved the correctness from 20% to above 50%

Integration and Application of Regulatory-Metabolic Network

December 1, 2014 – May 1, 2015

· Built up an integrated metabolic-regulatory network for yeast based on a new automatic metabolic-regulatory integration algorithm, EGRIN-PROM · Proved a strong phenotype prediction ability of EGRIN-PROM integrated network, by comparing the Matthews correlation coefficient (mcc) with YEASTRACT-PROM network · Performed optimization and simulation of the flux in yeast 7.00 metabolic network, aiming to improve the expression of acetoacetyl CoA · Proposed gene modification strategies including gene knockout and gene over-expression that could improve the expression of acetoacetyl CoA · Proves the accuracy of modification strategies based on pathway analysis

Synthetic Biology Software Design-“EASYBBK”

December 1, 2013 – October 1, 2014

· Participated in functional design of “EASYBBK” – an assistant tool achieving evaluation, visualization and simplification of Biobricks by requirements survey · Designed an assessment model of biobricks based on its status, reliability, feedback and relevant publication · Integrated relevant information and data of all extant biobricks in International Genetically Engineered Machine (iGEM) official Registry · Achieved sequence alignment function against our own database using NCBI Stand-alone BLAST · Won a gold medal in iGEM competition as an EASYBBK software team member Project Website: http://2014.igem.org/Team:SJTU-Software

Mechanism Exploration of Selectivity of PKB Inhibitor

October 1, 2012 – July 1, 2014

· Explored the selectivity mechanism for protein kinase B inhibitors with molecular dynamics simulation(MD) · Conducted 3D-QSAR modeling on PKB inhibitors and PKA inhibitors thus proving its high prediction ability · Revealed possible methods to improve the selectivity of inhibitors based on simulation results