I lead the product analytics vertical at Pluang, a multi-asset investment app for Indonesians. I do a mix of product, engineering and customer experience.
I enjoy writing. It helps me think clearly and speak with confidence in bigger rooms. Once a year, I go on a solo trip.
My values:
Keep my promises and show up for people
Start from the customer's problem
Have good taste, sweat the details and take pride in my work
Aim to be the best and get there by iterating fast
Help the people around me grow
QueryGPT
Problem
90% of business users couldn’t write SQL. Routine data questions waited hours for the Data team to answer.
Solution
Built, maintained and integrated the context layer into a third-party AI analyst. It includes metric definitions, business logic and sample queries.
Result
Saved the Data team over 700 hours in 3 months. QueryGPT is now used daily by the C-suite, Finance, Product and Growth teams.
+Details
That is roughly the work of 2 full-time analysts. Building the context layer and keeping it up to date with the latest business updates and product releases took the most effort. Now, QueryGPT has an accuracy of 85%, and routine queries to the Data team have fallen by 70%.
Events AI Pipeline
Problem
PMs tagged every button click and page view manually from Figma, and only 60% of events passed QA.
Solution
Built an AI pipeline that reads the PRD and designs, checks the existing schema and drafts each feature’s events for PMs to approve.
Result
Over 95% of events now pass QA, so the Product and Design teams can self-serve analytics with AI agents.
+Details
Event instrumentation work per feature fell from 10 days to 2.5 days when we piloted it on 7 new features in the Q2 2026 release. Now, it takes less than a day to ship those events to production. I created a reusable playbook to onboard every PM within 2 weeks, and removed the Data team from the QA cycle.
Platform Retention
Problem
Half of new users who deposited money left the platform within a year.
Solution
Analysed these users to profile who churns early, and proposed product changes to the top-up and transaction funnels to retain them.
Result
Users who start with US stocks are about 30% more likely to stay after a year than those who start with crypto. This shaped a key retention KPI: a second purchase within 30 days.
+Details
The analysis covered 535K users. After a year, 66% of US-stock-first users stay, against 51% of crypto-first users. Interestingly, users who deposit over $100 in their first month also retain better. I presented the findings to the CEO, COO and VP of Product.
Low CSAT Framework
Problem
A quarter of rated support tickets got 1 or 2 stars for customer satisfaction (CSAT), and no team could connect the platform issues, service quality and owners behind them.
Solution
Reviewed 100 low-rated tickets and pieced together each customer’s full journey to see why they rated us poorly.
Result
67% traced back to four broken funnels: account security, KYC, deposits and withdrawals. Reporting to the COO, I led the team to create a new SOP for human agents and an FAQ knowledge base for the chatbot.
+Details
I built a driver tree that splits low CSAT into product root causes and CS handling failures, with an owner team for every branch. 80% of tickets were also mishandled, mostly through templated or delayed replies from human agents. I now lead the redesign of the CS framework with the new AI enablement engineer and Senior CS Manager.
Seniors couldn’t read their government letters, and eldercare centre staff spent up to 2 hours on each one.
Solution
Seniors told us they’d rather use their phones than a kiosk, so we built a WhatsApp bot that turns a scanned letter into a plain-language summary and reads it aloud in 4 languages and 2 Chinese dialects.
Result
A routine case went down to about 30 minutes. Empower Ageing, a charity for seniors, now carries the service on.
+Details
I secured SGD 20,000 to pilot it at Bukit Panjang Community Club and led all 5 partners, including the People’s Association, Fei Yue Community Services and ElevenLabs. It processed 1,450 letters across 13 types, from housing to hospital notices. I learned to take a product from 0 to 1 and pitch it on stage at the Build for Good finale, and it was covered in national newspapers, on radio and on TV (Mandarin interview).
Organised a hackathon with GoodHub SEA to tackle the problems of an ageing society. Caregivers over 60 were vibe-coding alongside student engineers.
Note Buddy
Led our final-year team on an early POC of SingHealth’s clinical digital scribe tool. Voice models were not that advanced then, so the hard part was separating speakers who talked over each other and switched languages mid-sentence. Note Buddy is now used by thousands of healthcare staff in Singapore.
Container Load Optimizer
Solved a supply chain optimisation problem in Python at Procter & Gamble, packing containers fuller than the commercial software on 9 of 10 test days. Learned how to read math papers and turn them into code. Realised I prefer working with people more than computers, and that good documentation matters.
Student Exchange
Spent half a year living in a culture (and vibe) completely different from Asia's. Travelled around the US and watched my first live NBA game.
Resident Assistant
My fondest memories of college, and my first home in Singapore. Organised events that brought local and international students together.
