Summary

Highlight: Data scientist/Machine Learning Engineer with 5+ years’ experience building end-to-end data apps; data pipeline, modelling, continuous re-training pipeline to end deployment

Mathematical/Statistical Skills

Derivation: Advance calculus of differential equations, stochastic calculus and real analysis

Modeling: Machine learning, time series analysis, sampling, ANOVA analysis, A/B testing, linear algebra, model/feature selection

Hands on experience: applying several ML/statistical algorithms to real world problems: (deep) Neural Network, Recurrent Neural Network, Gradient Boosting, Ensembling techniques (inclusive of stacking and blending), Clustering, GLM, Markov, Game Theory and Simulation Models.

Programming Skills

Functional programming: Python, PySpark

Distributed systems: Hadoop, Hive, Spark

Databases: MySQL, Postgres, Redshift

Data visualization: R (inclusive of Rshiny), Python, Tableau

Model deployment/Orchestration: Airflow, Docker, Kubernetes, MLflow

Tech Stack: AWS

Professional experience

Senior Machine Learning Engineer

ShopBack

Sep '20 to Present

Developing and productionize deep (and machine) learning recommendation models in-house

Responsible for end-to-end MLOps

Data pipeline and model development

Continuous re-training of model

Model deployment, maintenance and AB testing

Productionized models

Deals recommendation for offline vertical:

Designed and built a learning to rank (LTR) model from gradient boosted trees

MLflow is used as the model tracking and registry

Batch prediction of a daily update

Daily airflow job that predicts user cohort recommendation, and uploads results to Postgres

FastAPI is dockerized and deployed onto (Kubernetes) EKS cluster

AB tested against manual curation, with model beating manual convincingly

Automated weekly retraining pipeline that decides if the retrain model will replace current production model by comparing evaluation metric (MRR) on same test dataset

Merchant recommendation for online vertical:

Designed and built a rNN deep learning model (GRU4REC) from scratch

MLflow is used as the model tracking and registry

Real-time prediction

Model client service (FastAPI) is dockerized and deployed onto (Kubernetes) EKS cluster

Model serving (BentoML) is dockerized and deployed onto (Kubernetes) EKS cluste

AB tested against AWS personalize model, and new model managed to beat AWS comfortably

Automated weekly retraining pipeline, that decides if the retrain model will replace current production model by comparing evaluation (MRR) metric

Productionized model result

Deals recommendation for (SBGO) offline vertical: +~15% in CTR and +~20% in transactions

Merchant recommendation for (SBOC) online vertical: +~3% in CTR and +~4% in transactions

Data Scientist/Founder

UpTick (Data Consultancy)

Dec '18 to Present

Providing data (science and engineering) consultancy services to SME and start-ups

Working with clients and built out end to end data products, notable clients are:

Ablr (fintech)

Built out ETL and data infrastructure

Created and integrated a credit (score) engine that leverages on 2 deep neural network (dNN) models of loan default and will user be late for payments

Created and integrated a differential interest rate pricing model for loans with differing credit score

3PlayGrounds

Built out ETL and data infrastructure as a baseline for having more robust and useable data for future developments like predictive analytics

GrowthOS

Creating content that is related to data analysis and (pirate) growth metrics

Data Scientist 2

Skyscanner Ltd

Mar '20 to Aug '20

Attached to Skyscanner accommodation ranking team

Worked on hotels ranking problem of returning the best sorted search results personalized to the user and search parameters

Implemented new features and model (LighGBM) that improved hotel booking relatively by 10%

Created a dNN model to replace current model, offline evaluations improve model’s MRR +~5%

Data Scientist 2

Skyscanner Ltd

Oct '18 to Mar '20

Attached to Skyscanner car hire vertical

Integrated an improved cross-sell model for car hire with deep learning that is more scalable with better evaluation metrics that has increased car hire searches and redirects by 5% and 1% respectively

Integrated a lead time model of when is the best time interval to send a push notification for the car hire cross-sell model

Created and integrated a price prediction model for car hire itinerary, with 80% accuracy (using RMSE)

Created a vertical recommendation, decision to recommend car hire or airport transfer for different cities

Created and tested a result sorting algorithm for car hire search results page, which improves redirects and bookings by 3% and 1% respectively.

Growth

Skyscanner Ltd

Apr '17 to Oct '18

Developed 2 in-house models from scratch for Multi Touch Attribution

Created a predictive model for 7 days retention as part of early warning system for business metrics

Improved app onboarding experience by leveraging on ML to determine what app features higher retained users are using in their initial month of installs

Maintained and improved a statistical self-servicing tool that proves causal impact on changes made to Skyscanner.

Data Analyst

Firemonkeys Electronic Arts

Mar '15 to Mar '17

Involved in scoping of AB tests and conducted statistical analysis on test results to provide insights and recommendation

Created simulation models for game events to validate and assist the tuning of values from product managers. Examples: Time taken, or total spend to complete events

Created and tested a price optimization model that has increased weekend revenue by 80%

Developed a bottom up forecasting revenue model that is now used by EA company-wide

Created a user acquisition spend optimization model that minimizes risk and maximize returns that has helped acquisition team for their buy strategy and direction

Managed an associate analyst and trained up his technical and soft skills to be a data analyst

Awards/Side Projects

Sky High Award

Skyscanner Ltd

Sep '19

Built out data products for other teams outside of car hire

Successfully mentored a marketing graduate to transit into a Data Analyst role

Speaker at AnalytiCon

Firemonkeys Electronic Arts

Aug '16

Presented as the last speaker on day 1 in EA inaugural Analytics Conference on content planning saturation using the statistical theory of mean reversion

Talent Spotlight Award

Firemonkeys Electronic Arts

Feb '16

Recognized and mentioned in EA Mobile quarterly general meeting for building the bottom up forecasting revenue model for company-wide use

Education

Major in Quantitative Finance

Singapore Management University

Aug '18' to Dec '14

Personal Details

Nationality

Singapore

Working rights in Australia

Requires sponsorship

Languages

English (native)

Mandarin