Writing
In-depth writing on data engineering, quantitative finance, Islamic capital markets, and building production data systems in financial services.
Herding behavior when investors abandon their own analysis to follow the crowd is measurable, predictable, and more structurally pronounced in Islamic equity markets due to a smaller eligible universe and homogeneous investor base. Neural networks and sentiment analysis detect it earlier than classical regression. Here's the full framework, the code, and a complete backtested scenario.
Read →Standard VaR assumes normally distributed returns. Islamic equity portfolios structurally screened, lower-leverage, sector-concentrated have return distributions that are decidedly non-normal. The Cornish-Fisher expansion within a four-moment framework gives you a more honest picture of tail risk. Here's the mathematics, a full Python implementation, and what it reveals about Islamic equity indices under market stress.
Read →The Capital Asset Pricing Model has a foundational riba problem its risk-free rate is built on interest. The Maqasid al-Shariah Compliance Asset Pricing Model replaces that foundation with Islamic wealth distribution logic and investor sentiment. Here's the mathematics, the implications, and a Python implementation.
Read →Classical portfolio theory assumes probabilistic risk but Shariah compliance introduces hard binary constraints that break standard optimizers. A genetic algorithm approach using fuzzy semi-spreads offers a principled alternative. Here's why it matters, and how it works.
Read →Shariah screening applies hard binary thresholds 33% on debt ratios, 33% on cash and receivables, 5% on haram revenue. These aren't soft guidelines. They're algorithmic cutoffs that restructure the investable universe and statistically reshape portfolio yield and volatility. Here's the data, the code, and what it means for fund managers.
Read →A deep dive into TurboQuant a new vector quantization algorithm from Google Research that achieves near-optimal distortion rates with virtually zero indexing overhead. And why I think it matters far beyond LLMs.
Read →While the AI industry races to make models bigger, CERN is burning microscopic neural networks directly into silicon to filter 40,000 exabytes of particle collision data per year in under 50 nanoseconds. The lessons reach far beyond physics.
Read →A deep technical guide to Monte Carlo simulation from the mathematics of random sampling and variance reduction to practical Python implementations in quantitative finance, insurance pricing, and risk modelling. With code.
Read →Most data governance initiatives fail not because of bad technology, but because they're treated as a one-time implementation rather than an ongoing discipline. Here's what actually works in a regulated financial services environment.
Read →After six months of using Polars in production pipelines, here's an honest assessment.
Read →Slowly changing dimensions are a solved problem but most implementations lean on Spark or SQL. Here's how to do it cleanly in Polars with about 60 lines of Python.
Read →Bronze, silver, gold the medallion pattern is everywhere. Here's what it actually solves, and where people get it wrong.
Read →Medical and takaful claims data is messy, late-arriving, and regulated. Here's how the bronze-silver-gold pattern maps to the reality of an insurance data platform.
Read →Delta Lake's time travel feature is genuinely useful for data reprocessing and debugging. Here's how to use it well and what it doesn't cover.
Read →After years of hiring data engineers in Malaysian financial services, here's the honest version not the job description version.
Read →The business case for migrating off SSRS is real. The migration itself is more nuanced than a licensing slide deck suggests. Here's what to prepare for.
Read →A short note on why I decided to write more, think out loud, and share what I build.
Read →