Malaysia
Interested in Fintech and Big Tech. Joined MoneyLion as a Data Scientist since June 2023. Previously a fresh graduate majoring in Data Science & Analytics, the university taught me hard skills like programming, calculus, probability, linear algebra, statistical analysis, optimization, simulation, database, and deep learning. I've also done a few internships in the domains of e-commerce, augmented reality, and oil & gas. In the future, I would probably still want to work with analytics and machine learning, and work on a business of my own. Apart from work, I enjoy travelling, piano, some table tennis and RTS video games.
Research & POCs • Custom Transaction Data Labeling Systems • Custom Model Monitoring Techniques • Counterfactual Learning & Explanation Frameworks ML System Consolidation and Integration • Created & led ML model consolidation strategies to reduce maintenance costs, eliminate redundant logic, and improve development effectiveness & efficiency. • Created & led use case integration (between MoneyLion, LifeLock, and NortonMoney) to enable any use case to plug into the same Transaction Enrichment foundation. People Manager in-Training • Instead of just owning project growth, practicing to own people's career growth as well.
Transaction Categorization (e.g. Groceries, Restaurants, Automotive) for: • PFM (e.g. track spendings, budgeting) • Financial Behavior Understanding (e.g. user persona) • Competitor Analysis (e.g. what other financial products are our users using) • UX Enhancement (e.g. tailored onboarding flow) ML Model Governance • Custom analytical tools (e.g. data segmentation) • Custom ML Frameworks (e.g. feedback loops, retraining pipelines) • Custom ML Templates (e.g. architecture-specific modeling codes) • Pre-production Model Audits
• Developed an Android app for identity recognition with face and ear biometrics that takes in simulated environmental context as a proof-of-concept (POC) using Kotlin, MLKit and AWS (S3) • Trained ear detectors with transfer learning using YOLOv5s, EfficientDet, FiftyOne and TFLite • Hosted recognition server using Docker, Flask and AWS (API Gateway, Lambda, Sagemaker, EC2, ECR)
• Ran advertising campaign with Facebook, Instagram and Reddit, captured around 20 newsletter subscribers in the first month • Set up data pipeline to ingest data collected from Google, Facebook and Reddit analytics using Fivetran and Snowflake, and performed customer segmentation using metrics such as acquisition and engagement • Conducted market research, user interviews and prior-art search to derive solution for Healthcare InnoMatch 2022 with colleagues, successfully advancing into 3 out of 4 rounds of evaluation • Collaborated with colleagues to deduce core technologies required to future-proof the company’s augmented reality (AR) smart glasses • Curated an industrial design checklist to ensure the the company’s smart glasses’ design adhere to human-computer interaction (HCI) design principles
• Worked in a team of 3 to establish WinSight to connect pre-u students with seniors, engaging more than 80 users upon launching • Built and maintain its database and website using HTML, CSS, JavaScript, Bootstrap, Heroku, Python, Flask and PostgreSQL • Reached out to users online and physically at Sunway University to obtain feedback and promote the platform WinSight connects pre-university students with university students as mentors to help them with everything from applications to financial aid to settling in - all completely free through our platform. Check out WinSight at winsight-dev.onrender.com!
• Liaised with the logistic department to build a linear regression model for delivery time prediction using R, achieving 0.56 in adjusted-𝑅2 • Reported modeling methods and findings to senior data scientists, as well as presenting them to the relevant stakeholders in business terms • Designed PowerBI dashboards to track business benefit of gas pricing model
• Participated in a team of 4 to detect e-commerce promotional fraud, attaining 61% in test precision • Analyzed large data sets provided by VNLife using Python, PySpark and NetworkX • Implemented fraud detection algorithms based on bipartite graphs from academic literatures, and derived new ones based on graph size, degree centrality and other properties, boosting results by 68%