Singapore, Singapore
Lead and manage the Data Science Risk, Security and Fintech team. Deliver impact to users in and outside the GoTo eco-system.
Building and leading the data science teams of 8 with focus on marketplace risk and credit risk management and fintech applications. Some of the key results are 1. Tokoscore Tokoscore is the pineering alternative credit score in Indonesia and the flagship product of Semangat Digital Bangsa (SDB), developed based on ecommerce data. Our team have been continuously iterating Tokoscores to provide the best performance to our internal parterns across GoTo ecosystem and external clients from major fintech players and banks in Indonesia. Besides Tokoscore, I also lead the team in developing other data products, including address validation, phone validation, income prediction, etc, and actively involve in sales event of SDB. 2. General anomaly detection system (GADS) Our team work with the domain experts in Tokopedia risk management team to craft a set of 20+ features to capture the characteristics of transactional collusion fraud that abuses our incentives or does cash advance. A cutting edge outlier detection algorithm is tuned to detect suspicious/abnormal shops in clusters through offline validation and iterations. GADS in production contribute to substantial subsidy saved daily and improve the overall loss saved by the whole anti-fraud system by a wide margin. 3. Counterfeit detection model Designed and developed a model that actively flags fake vitamin products in Tokopedia. It combines features from free text data like product reviews, and structural shop and product level characteristics according to domain knowledge. The developed model has superb performance in detecting related counterfeits, achieving 90%+ precision and 80% recall. 4. Loan application propensity model Propensity models that predict the users' likelihood to apply for a certain loan products. It helps marketing teams in both Tokopedia and GTF in lead generation and it has shown to double the CTR/CVR in various communication channel such as in app push notification and WhatsApp.
Lead 2 data scientists for merchant lending business in Tokopedia and GoBiz. Our scopes include 1. Application and Behavior scores for ModalToko and GoModal, which are cash loans products for Tokopedia and GoBiz merchants, respectively. 2. Develop real time underwriting scores and services that utilize data of users' application information and credit bureau reports. 3. Merchant GMV prediction models that facilitate merchant lending risk team in underwriting and account management 4. Merchant archetypes: a set of merchant profiling/segmentation models for business team to tailor their loan products and GTM strategies to our merchants' characteristics.
Lead a multi-disciplinary team of 30, including algorithm engineers (data scientists), BE/FE engineers, product managers and data analysts. Invented, developed, and scaled Shopee and SeaMoney's in-house eKYC solutions from scratches, which includes IC card OCR, IC anti-spoofing, face recognition, quality check of KYC photos as well as selfie anti-spoofing (liveness check) that is deployed in both edge devices and Sea's on-premise clusters across SEA and LATAM. The eKYC services we built has been serving millions of users for all Fintech and e-commerce business across the globe, with superb reliability and state-of-the-art performance in the regions.
Acted as project lead for a few new initiatives in Shopee's data science department, including fake new accounts detection, SPU matching and clustering, as well as building Shopee's first KTP (Indonesian IC card) OCR models
Worked both independently and collaboratively on a wide portfolio of projects, including SKU grouping, sales forecasting, traffic forecasting
Provide tutorial and guidance to students of Modules EE2024 Programming for Computer Interfaces, EE4302 Advanced Control Systems, EE4307 Control Systems Design and Simulation and EE5107 Optimal Control System
Advising undergraduate students on Final Year Projects