Pricing uncertainty with modern machine learning.
I'm HyunSu Kim. I build tools and models where actuarial thinking meets AI — and I teach predictive analytics from first principles.
Projects at the intersection.
Experiments and tools exploring what machine learning can do for insurance — and what insurance can teach machine learning about risk.
Claim severity, beyond GLMs
Benchmarking gradient boosting against classical GLM pricing on open auto-insurance data, with SHAP-based explanations an underwriter can actually read.
Read the write-up → NLP · LLMPolicy documents, parsed
An LLM pipeline that extracts coverages, exclusions and limits from policy wordings — turning PDFs into structured, queryable data.
Read the write-up → ForecastingMortality, forecast forward
Comparing Lee–Carter with neural approaches for mortality forecasting, and what the gap means for long-term products.
Read the write-up →Predictive analytics, taught from first principles.
A course for actuaries, analysts and the actuarially curious — built around real insurance problems, not toy datasets.
- From GLMs to gradient boosting: what changes, what doesn't
- Model validation the regulator-friendly way
- Hands-on: pricing, reserving and lapse models in Python
First cohort in production
Leave your email and be the first to know when enrollment opens. No spam — one email when it's ready.
Actuarial thinking, modern tools.
I've spent my career in insurance and can't stop tinkering with machine learning. This site is where the two meet: personal projects, notes, and a course on predictive analytics.
I write and teach in English, work from Seoul, and am always happy to talk about pricing, reserving, or why your model's lift chart looks suspicious.