Building Analytics Services for Actuaries and Underwriters With FastAPI
Why actuarial logic belongs in a service, how to structure one with FastAPI and Pydantic, test it properly, version it, and put it in...
Writing
Notes on actuarial analytics & data
Articles on risk modelling, data engineering and the London specialty market, with code you can run.
Working With Microsoft Fabric for Actuarial Data: Lessons From Practice
Lakehouse or warehouse, medallion layers, idempotent PySpark merges, Delta table maintenance, Git integration and capacity: practical lessons for running actuarial pipelines on Microsoft Fabric....
Data Modelling for Insurance Analytics: Grain, Stars and Time
A dimensional model for premium and claims, why transaction facts beat snapshots of totals, slowly changing dimensions, and how to build reserving triangles from...
Working With Data in the London Insurance Market: A Field Guide
Lines, layers, line sizes, bordereaux, currencies and dates: the concepts that trip up analysts new to London market data, and how to handle them....
Machine Learning in Insurance Pricing: Where It Helps, Where GLMs Still Win
A practical view of GLMs and gradient boosting for insurance pricing: Tweedie losses, exposure offsets, explainability, and why specialty lines need a different approach....
Data Engineering for Insurance Analytics: Principles That Survive Contact With Bordereaux
Idempotent loads, explicit grain, validation at the boundary, reconciliation and lineage: the data engineering habits that make actuarial and underwriting numbers trustworthy.
Copulas for Insurance Dependence: A Practical Guide and Its Pitfalls
Why correlation is not enough for tail risk, how Gaussian and t copulas differ, how to simulate and calibrate them in Python, and the...
Monte Carlo Loss Simulation for Specialty Insurance, From First Principles
Build a frequency and severity simulation in NumPy, apply an excess of loss layer, read OEP and AEP curves correctly, and know how many...
Emerging Risks in the London Specialty Market: What the Data Can and Can't Tell Us
Cyber, war and political violence, secondary natural perils and accumulation: how emerging risks show up in London market data, and why analysts should be...