Senior Consultant at EY. I build behavioural models for interest-rate risk in the banking book, on deposits and financing.
Further Mathematics, Mathematics, Physics, and Chemistry.
Set out for mechanical engineering, then turned to econometrics.
Machine learning applied to financial forecasting.
NFT study for the Hong Kong e-wallet.
I'm a Senior Consultant in Financial Risk Management at Ernst & Young in Hong Kong. I build behavioural models for interest-rate risk in the banking book: how non-maturity deposits, term deposits, and financing respond when rates move, for banks in Hong Kong and Malaysia.
I grew up in Malaysia and started toward mechanical engineering before turning to econometrics. That path ran through a BSc in Econometrics and Data Science at the University of Amsterdam and an MSc in Financial Technology at HKUST, and then to EY.
Outside of work I follow commercial aviation — airliners, airport operations, and the way flight networks fit together. It keeps reminding me how much of a large system depends on precision.
Senior roles in financial risk, model validation, and AI-driven solutions at leading institutions.
Specialised in advanced machine learning and application to various domains such as financial forecasting. Covering all topics such as deep learning, graph neural networks, algorithmic trading, blockchain.
Rigorous quantitative training combining econometrics, causal inference, time-series analysis, and modern data science methods
Foundational AI curriculum covering symbolic AI, Logic, Algorithms and Data Structures, Programing in C
Achieved straight A* in Further Mathematics, Mathematics, Physics, and Chemistry. Initially focused on mechanical engineering but developed a strong passion for data science and AI, leading to a pivot in academic focus.
Small tools I actually use. More will land here as I ship them.
Paste a URL or a whole message. I read the address in your browser — lookalike names, hidden destinations, and how new the domain is. Nothing is sent to this site, so it doesn’t spend the Vercel allowance. HTTPS only means the connection is encrypted.
Drop a QR screenshot, paste the LPA string, or type the SM-DP+ fields. I build a link you open on the phone that needs the plan — iPhone or Android does the rest. Nothing is uploaded.
A public tester of a shared clipboard. This copy asks you to press Pull, so an open tab does not spend the Vercel allowance. The GitHub project still checks for new slips every 1.5 seconds. Email me for a trial code.
Terminals, aircraft, ground vehicles and trains, running in your browser. Opening it downloads the scene — about 10 MB, and more if you fly over to the circuit. Nothing is sent to this site, so it doesn’t spend the Vercel allowance.
The pairwise interpretation of ROC AUC — 'probability a random positive outranks a random negative' — isn't just a nice intuition pump that happens to give the same number as the curve. It's mathematically the same computation. Here's why, worked through an 8-point example both ways.
Fraud, default, churn, cancer — the outcome you actually care about is almost always the rare one. SMOTE and class weights fix the training problem, but they quietly break your predicted probabilities unless you correct the intercept back afterward. Here's the fix, from King & Zeng (2001).
Wilson confidence intervals are a great governance tool for monitoring behavioural assumptions — but at large sample sizes they get so tight that trivial, immaterial drift starts failing the check. Cohen's h fixes that by measuring the size of the gap, not just whether it's detectable.
Fitting a model by maximum likelihood almost always means calling scipy.optimize under the hood. BFGS, L-BFGS-B, Nelder-Mead, Powell, and basin-hopping all claim to find 'the' maximum likelihood estimate — but they make very different assumptions, and picking the wrong one silently hands you a local optimum instead of the answer you wanted.
In 1956, Edsger Dijkstra designed an algorithm in his head, in 20 minutes, without pen or paper. It still powers Google Maps, internet routing, and game AI today. A complete guide to the mechanics, intuition, and limitations of one of computing's most elegant ideas.
Static behavioural assumptions power ALM and IRRBB models, but they drift. Wilson score intervals give you a statistically sound, auditable way to monitor when they need recalibration — without building a full dynamic model.
When market rates move, customer rates rarely follow one-for-one. Pass-through rates explain the gap, the timing, and the IRRBB risk hiding inside margin forecasts.
A beginner-to-fluent guide to Polars LazyFrame and PySpark. Same verbs, two spellings: filters, columns, joins, windows, strings, dates, and the null traps that actually change results.
After years of being the default, Pandas is hitting its limits on modern data lakehouse workloads. Here's why I switched to Polars + PyIceberg — and why you might want to as well.
Investment Opportunities in the Gaming and eSports Industry
Hong Kong University of Science & Technology
Bachelor Thesis
University of Amsterdam
Assignment: High Bias in Labour Market
University of Amsterdam
Empirical Project
University of Amsterdam
Colour in Visual Cognition
University of Groningen
I’m always open to discussing interesting AI & finance problems, consulting engagements, research collaborations, or just connecting with fellow fintech enthusiasts.
Drop me a message and I’ll get back to you within 48 hours.
