01 — The work
A good model helps answer a question. A good product makes that answer useful. The practice connects the analysis with the experience.
The projects range from healthcare market analysis and financial research to native applications and websites. They share the same aim: make complicated information easier to understand, explore and act on.
02 — Disciplines
Different questions.
Connected skills.
Actuarial and finance studies inform the treatment of risk and uncertainty. Software makes those ideas testable and explorable. Each project adds something to the next.
-
Machine learning
Sensor fusion · multimodal · retrieval
Turning sensor data, documents and messy records into something useful. The work starts with understanding the data, then building and testing a model around the real constraints.
-
Quantitative finance
Execution · risk · backtesting
Research tools, trading workspaces and risk models that make assumptions easier to inspect. From exploring a strategy to understanding how it behaves when conditions change.
-
Actuarial studies
Loss models · reserving · health economics
Making sense of demand, cost and risk in healthcare and insurance. Practical models for comparing scenarios, testing assumptions and planning the next decision.
-
Software & data
Pipelines · scale-to-zero · tooling
The connections that make a useful product possible: reliable pipelines, native applications, reporting tools and websites built around how people actually work.
-
Web & product design
Websites · interfaces · interaction
Websites and product experiences that make the next step clear. Thoughtful layouts, purposeful motion and the engineering to make them work across devices.
03 — Selected work
A few places to start.
- 01 Ichne Learning what wrist motion reveals about a shift. Wearable research · Activity recognition 2026
- 02 Chronos Research, test and inspect a trading idea in one place. Trading research · Desktop applications 2026
- 03 Aphelion Make the first draft. Keep control of every detail. Native design software · Local AI 2026
- 04 Health Index A clearer view of where healthcare services are needed. Health economics · Geospatial ML 2025
- 05 Specialist Projections Explore how a changing population affects specialist demand. Demand modelling · Health workforce 2025
- 06 Radiology Workforce Model Turn imaging activity into reviewable workforce estimates. Workforce modelling · Health economics 2026
04 — Recently added
The next chapter.
Recent applications and research, building on the data tools and models that came before them.
05 — Research & writing
Room for a
longer explanation.
-
Transformers for Market Prediction
A first-principles course connecting neural networks and transformer architecture to the mathematics of forecasting, evaluation and trading research.
Read the overview -
Financial Markets: A First-Principles Course
A guide to the markets around a prediction model: participants, order books, transaction costs, financing and the ways trading systems can fail.
Read the overview -
Making Healthcare Markets Comparable
How to compare healthcare markets when their payment systems, boundaries and published statistics do not line up neatly.
Read the overview
22
Projects to explore
5
Connected disciplines
7
Books & research notes
01
Independent practice
06 — Approach
From the first question
to a working tool.
-
01
Understand the problem
Start with the decision, the people making it and what a useful answer would look like.
-
02
Get close to the data
Check the sources, find the gaps and work out what can be established before choosing a model.
-
03
Build something useful
Bring the analysis into a tool people can work with, from a focused calculator to a complete application.
-
04
Test and improve
Compare the output with a baseline, explain the limits and use what was learned to shape the next version.
The details change from project to project. The care behind them stays the same. More about the practice →
07 — What matters
Care in the work.
Clarity in the result.
Useful work needs clear reasoning and careful execution. These principles guide the models, applications and research across the practice.
-
01
Make it checkable
Keep the source close to the result. Someone else should be able to follow the reasoning and see where a number came from.
-
02
Be clear about the unknowns
Show which parts are measured, estimated or still being tested. That makes a model more useful to the person making the decision.
-
03
Choose the right tool
Use dependable calculations for the numbers and machine learning where it helps. The simplest approach that solves the problem is a good place to start.
-
04
Learn from what failed
Keep a record of the dead ends as well as the wins. They explain the choices in the current build and make the next version better.
08 — Work together
Something you’d like
to figure out?
Start a conversation about a product to build, a dataset to make sense of or a research question worth exploring.