Research & writing · 7 pieces
The thinking
behind the work.
Books, working papers and notes from the questions that keep coming up in practice. How models learn. How markets behave. How to make a decision when the data is incomplete.
A growing body of work
Ideas become something you can revisit.
A library of connected volumes: each piece holds its own argument, with paths back to the ideas and methods around it.
Layered volumes and travelling traces form an abstract research library. The illustration is conceptual.
Spatial studyBrowse the research library
-
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 -
Counting a Market Without a Single Register
A study of how accreditation records, establishment data and service activity can be brought together to estimate a healthcare segment.
Read the overview -
Evaluating Local One-to-Many Identification
A design and evaluation study of local identification, covering gallery size, image quality, label noise and the choice between automatic decisions and human review.
Read the overview -
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 -
The Programme: An Operational Playbook
A step-by-step companion to the market-prediction course, turning the research programme into concrete actions, saved outputs and decisions made before results arrive.
Read the overview -
How Small Transformer Models Work
An illustrated introduction to the operations inside a transformer, from vectors and attention to training and inference, using language and wrist-sensor examples.
Read the overview
Each entry describes its scope and stage of development. Client research is presented as an anonymous methods overview.