resume

Bachelor of Computing, Computer Science

National University of Singapore

• Special Programmes - Turing Research Programme, University Scholars Programme (Honour Roll)
• Relevant Coursework - Machine Learning (Top Student, Best Project), Computer Vision (Best Project), AI Planning & Decision Making (Best Project), Natural Language Processing, Data Structures & Algorithms, Software Engineering, Research Methodology

Data Scientist Intern

Indeed - Match Recommendation Platform

• Feature Experimentation - Improved Indeed's core recommendation system to match jobseekers and employers, by experimenting on a new "no. of hires needed" feature, improving job outcomes for >300M jobseekers.
• Data Engineering - Added new features of >300M jobseekers of >100 days to training database, backfilling historical data over aggregated windows using Spark's reduce(). Orchestrated data dumping to AWS servers job using Airflow.
• Feature Analysis - Performed EDA and feature engineering on new features. Trained and evaluated 4 new models against original models using internal ML platform, showing improved AUROC and justifying a full A/B experiment.
• A/B Experimentation - Performed power analysis to determine required traffic size and ran A/B experiments using Proctor with orthogonal design to account for interaction effects. Deployed improved models to production.

Student AI Researcher

NUS - Collaborative, Learning, and Adaptive Robots Lab

• Transformer Models - Improved generalization of world models during distribution shift in the multi-agent context, by exploiting permutation invariance of transformers, exceeding performance of state-of-the-art Trajectory Transformer.
• Representation Learning - Improved robustness of reinforcement learning to noise in visual observations, by learning decorrelated latent variables using Barlow Twins, exceeding performance of state-of-the-art algorithm Dreamer.
• Survey Paper - Reviewed representation learning for model-based reinforcement learning, challenges and directions."

Machine Learning Engineer Intern

Grab - Marketplace

• Probabilistic Models - Built a probabilistic modelling framework used by Grab's core dynamic pricing algorithm, implementing 5 models using Tensorflow Probability (Gaussian mixture density network, Bayesian neural network, etc)
• Utility Tools - Built tools for evaluation on industry datasets (KL & JS divergence), visualization of model probability densities (3D and violin plots), and hyperparameter tuning using Ax Bayesian optimization framework.
• Internship Sharing - Presented to the data science community at Google Developer Space Singapore.

Executive (Machine Learning)

IMDA - Digital Services Lab

• Anomaly Detection - Designed an anomaly detection algorithm, achieving 0.92 F1 score and deployed as client's main solution, securing a $500K project deal, after evaluating >10 algorithms (e.g. LSTM, Holt-Winters, SARIMA).
• Natural Language Processing - (Confidential) Built hierarchical attention network with GloVe vectors using Keras and Spacy, scraped 40K webpages using Selenium, and delivered insights to Director of Tech & Infra Group.

Structured Multi-Agent World Models

• Researched graph neural networks for multi-agent reinforcement learning, improving performance and planning accuracy. Awarded CS4246 Class Project Competition Winner out of 142 students.

Barlow Twins for Model-Based Reinforcement Learning

• Researched Barlow Twins and contrastive self-supervised learning for model-based reinforcement learning, improving robustness and training stability.

Bayesian Multi-Agent Reinforcement Learning

• Researched flipout for competitive multi-agent reinforcement learning, improving performance, training stability and generalization. Awarded 1st Place at NUS Project Showcase out of 78 teams.

NLP Question Answering Transfer Learning

• Researched transfer learning of NLP question answering (QA) capabilities, benchmarking 6 BERT-based models over 14 QA datasets on zero-shot and fine-tuned performance.

• Honourable Mention Project - for CS4243 Computer Vision

• Top Project - out of 142 students for CS4246 AI Planning & Decision Making
• Silver (Achievement) - NUS School of Computing Student Awards
• 1st Runner Up - Optigram Data Visualization Competition
• Honour Roll - University Scholars Programme

• Top Student & Top Project - out of 272 students & 54 teams for CS3244 Machine Learning
• Top 10 Finalist - out of 54 teams at HackAsia Global Hackathon

• Best AI Hack - out of 130 participants at NTU iNTUition Hackathon

President at NUS Statistics and Data Science Society

• Led 40 students to organize data science workshops and an annual competition with 850 participants.

Technology Associate (Data Analytics) at Google Developer Student Club

• Organized 5 data science workshops at Google Developer Space Singapore, reaching 120 students.