chen - johntiger1.github.io

1
9/20/2019 Creddle | Resume https://resume.creddle.io 1/1 JOHN CHEN [email protected] johntiger1.github.io 4168398823 linkedin.com/in/johnn-chen/ Skills DATA SCIENCE AND DEEP LEARNING PyTorch scikit-learn PySpark Tensorow xgboost SparkSQL NumPy Pandas nltk SpaCy matplotlib Spark MLLib mlow FULL STACK PROGRAMMING Python Node.js AngularJS React JavaScript C Java Bootstrap PL/SQL Postgres MongoDB C# Racket HTML5 CSS3 Projects Reddit Political Alignment Text Classification Implemented character level language model; for text classication on Reddit posts. Implemented vanilla RNN as well as LSTM architecture and compared results against AdaBoost, SVM, logistic regression, Random Forest, and simple feedforward neural network. Implemented gradient clipping to deal with exploding gradients and vanishing gradients problem, improving F1 from 0.25 to 0.33. Education Master of Science in Computer Science Sept. 2019 to Current University of Toronto Research areas: Deep Learning and Natural Language Processing. Advised by Professor Frank Rudzicz. Thesis-based research degree, eligible for direct entry to PhD. Bachelor of Science in Computer Science Sept. 2014 to May 2019 University of Toronto 2019 Focus in Articial Intelligence and Deep Learning, Graduated with High Distinction. Cumulative GPA: 3.98/4.00 Employment Air Miles (a LoyaltyOne Company) Toronto, Ontario Data Science Intern May 2019 to Aug. 2019 Developed several machine learning models end-to-end, including ight redemption propensity model. FRP xgboost model achieved 0.97 AUC ROC, and has 9x lift compared to baseline. Improved model via seasonality adjustments, model-free feature selection algorithm and model-based genetic feature selection algorithm, as well as hyperparameter tuning via Bayesian Optimization using scikit-optimize. Model has been used in business and has delivered 40% e-mail cost savings in an acquisition campaign targeting over 500 000 customers. Performed research and development in recommendation engine prototype for leading Canadian grocery chain. Achieved 111% basket coverage improvement over baseline by using market basket analysis and frequency-based model. Vector Institute Toronto, Ontario Research Intern June 2018 to Current Developed and implemented several end-to-end deep learning solutions including: character-level recurrent neural network language model as part of industry partnership with DataX medical image captioning using convolutional neural network encoder and recurrent neural network decoder. Doubled F1 performance by reformulating problem as multilabel classication problem. paragraph-level style transfer in text between doctor and patient corpora (research ongoing) massively multilabel, highly imbalanced classication. Improved F1 performance by 156% over baseline pretrained network via image augmentation and netuning. University of Toronto Teaching Assistant Sept. 2016 to Current TA for 5 dierent courses: CSC209 (Software Tools and Systems Programming), CSC263 (Data Structures and Analysis), CSC165 (Mathematical Expression and Reasoning for Computer Science), CSC324 (Functional Programming and Principles of Programming Languages), and CSC420 (Computer Vision and Image Understanding) Ontario Teachers' Pension Plan Software Developer Intern, DevOps and Risk Analytics Sept. 2017 to May 2018 Improved deployment set-up time by 30% by developing full-stack RESTful C# Web API + AngularJS + Oracle database web app for managing UrbanCode agent and server registry information. Contributed to performance gains of 50x in nancial software through developing hybrid Powershell/C# solution to automatically generate, format and send e-mail reports to senior management daily. Centre for Computational Medicine at SickKids Hospital Research Student May 2017 to Aug. 2017 Awarded $6000 NSERC research grant to conduct summer research into Bayesian networks to improve automatic disease diagnosis algorithm used in Phenotips web platform. Sped up naive algorithm by 500x via dynamic programming and multithreading. Awards Vector Research Grant · Vector Institute June 2018 Awarded to conduct deep learning research under supervision of Dr. Frank Rudzicz 1st Place, Cisco DevNet HackathonCisco June 2016 Created an IoT app for real-time monitoring of plants, using Cisco’s Tropo VoIP service and Relayr API. Rotary Club of Toronto Community Champion Scholarship · Rotary Club of Toronto and Toronto Argonauts June 2014 Full tuition scholarship to University of Toronto

Upload: others

Post on 19-Oct-2021

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: CHEN - johntiger1.github.io

9/20/2019 Creddle | Resume

https://resume.creddle.io 1/1

JOHN [email protected]

johntiger1.github.io 4168398823 linkedin.com/in/johnn-chen/

SkillsDATA SCIENCE AND DEEPLEARNINGPyTorchscikit-learnPySparkTensor�owxgboostSparkSQLNumPyPandasnltkSpaCymatplotlibSpark MLLibml�ow

FULL STACK PROGRAMMINGPythonNode.jsAngularJSReactJavaScriptCJavaBootstrapPL/SQLPostgresMongoDBC#RacketHTML5CSS3

ProjectsReddit Political AlignmentText ClassificationImplemented character levellanguage model; for textclassi�cation on Reddit posts.Implemented vanilla RNN aswell as LSTM architecture andcompared resultsagainst AdaBoost, SVM,logistic regression, RandomForest, and simple feedforwardneural network. Implementedgradient clipping to deal withexploding gradients andvanishing gradientsproblem, improving F1 from0.25 to 0.33. 

EducationMaster of Science in Computer Science Sept. 2019 to CurrentUniversity of TorontoResearch areas: Deep Learning and Natural Language Processing. Advised by Professor Frank Rudzicz.Thesis-based research degree, eligible for direct entry to PhD.

Bachelor of Science in Computer Science Sept. 2014 to May 2019University of Toronto 2019Focus in Arti�cial Intelligence and Deep Learning, Graduated with High Distinction. Cumulative GPA: 3.98/4.00

Employment

Air Miles (a LoyaltyOne Company) Toronto, OntarioData Science Intern May 2019 to Aug. 2019

Developed several machine learning models end-to-end, including �ight redemption propensity model.FRP xgboost model achieved 0.97 AUC ROC, and has 9x lift compared to baseline. Improved model via seasonality adjustments, model-free feature selection algorithm and model-basedgenetic feature selection algorithm, as well as hyperparameter tuning via Bayesian Optimization usingscikit-optimize. Model has been used in business and has delivered 40% e-mail cost savings in an acquisition campaigntargeting over 500 000 customers.  Performed research and development in recommendation engine prototype for leading Canadian grocerychain. Achieved 111% basket coverage improvement over baseline by using market basket analysis andfrequency-based model. 

Vector Institute Toronto, OntarioResearch Intern June 2018 to CurrentDeveloped and implemented several end-to-end deep learning solutions including:

character-level recurrent neural network language model as part of industry partnership with DataXmedical image captioning using convolutional neural network encoder and recurrent neural networkdecoder.  Doubled F1 performance by reformulating problem as multilabel classi�cation problem. paragraph-level style transfer in text between doctor and patient corpora (research ongoing)massively multilabel, highly imbalanced classi�cation. Improved F1 performance by 156% over baselinepretrained network via image augmentation and �netuning. 

University of TorontoTeaching Assistant Sept. 2016 to Current

TA for 5 di�erent courses: CSC209 (Software Tools and Systems Programming),  CSC263 (Data Structuresand Analysis),  CSC165 (Mathematical Expression and Reasoning for Computer Science), CSC324(Functional Programming and Principles of Programming Languages),  and CSC420 (Computer Vision andImage Understanding)

Ontario Teachers' Pension PlanSoftware Developer Intern, DevOps and Risk Analytics Sept. 2017 to May 2018

Improved deployment set-up time by 30% by developing full-stack RESTful C# Web API + AngularJS +Oracle database web app for managing UrbanCode agent and server registry information. Contributed to performance gains of 50x in �nancial software through developing hybrid Powershell/C#solution to automatically generate, format and send e-mail reports to senior management daily. 

Centre for Computational Medicine at SickKids HospitalResearch Student May 2017 to Aug. 2017

Awarded $6000 NSERC research grant to conduct summer research into Bayesian networks to improveautomatic disease diagnosis algorithm used in Phenotips web platform. Sped up naive algorithm by 500xvia dynamic programming and multithreading.

AwardsVector Research Grant · Vector Institute June 2018Awarded to conduct deep learning research under supervision of Dr. Frank Rudzicz

1st Place, Cisco DevNet Hackathon⋅ Cisco June 2016Created an IoT app for real-time monitoring of plants, using Cisco’s Tropo VoIP service and Relayr API. 

Rotary Club of Toronto Community Champion Scholarship · Rotary Club of Toronto and Toronto Argonauts

June2014

Full tuition scholarship to University of Toronto