Selected engagements
Quantitative modelling and data engineering for enterprise clients across financial services, staffing, automotive SaaS, and public research. Each card opens a fuller writeup.
Adaptive regime-aware trading engine
Designing and building an adaptive multi-strategy trading engine for Complexa Labs: a regime-classification stack rooted in multifractal cascades and extreme value statistics feeds a portfolio of regime-gated strategies whose weights update online via Bayesian learning.
B2B customer segmentation & next-best-product recommender
End-to-end CLTV pipeline (mixture cure survival, quantile regression, GBM ensembles, calibrated A/B framework) followed by a causal state-space next-best-product recommender with cross-product effects identified via difference-in-differences and regression discontinuity.
Enterprise-scale NLP for unstructured text
Applied modern language-model techniques to extract structure and insight from unstructured text across a global automotive retail platform.
Macroeconomic scenario simulation framework
Non-linear constrained optimisation framework spanning 500+ interdependent variables across multiple planning horizons, integrating IPOPT, CMA-ES, and differential evolution solvers.
Pandas → Spark-native pipeline re-architecture
Re-architected a critical ingestion pipeline from pandas to Spark-native operations across a MongoDB/Azure environment, alongside production Python/PySpark microservices with FastAPI endpoints and Celery task orchestration.
Mortality burden of air pollution — epidemiological HIA
Modelled PM10/PM2.5/NO₂ effects on mortality under traffic scenarios via Poisson and panel fixed-effects models; applied causal forest and Double ML to air-quality perception survey data.