Code and experiments
Identifiability analysis for mathematical models
reparam
Julia code for Invariant Image Reparameterisation: finding identifiable parameter combinations and using them to reformulate models. The repository includes worked examples and profile-likelihood calculations.
Coding agents for scientific computing and teaching
I develop tools for working with coding agents on scientific code and documents, including shared interactive sessions and previews of mathematical writing.
Extensions for Pi
Pi Studio
A browser workspace for Pi, with prompts and responses alongside files and previews. You can comment on documents, edit prompts and work with a live REPL.
pi-markdown-preview
Preview Pi responses and local Markdown, LaTeX and code files in the terminal or browser, or export them to PDF. Browser previews can follow the session as you work.
pi-repl
Start a REPL that you and Pi share. Work at the prompt from another terminal while Pi uses the same variables and functions. Supported runtimes include Python, Julia and R.
Agent-agnostic tools
agent-repl
Work in the same running REPL as a coding agent. You can define variables and functions, ask the agent to use them, and inspect or change its work at the prompt.
agent-markdown-preview
Read coding-agent responses and local Markdown, LaTeX and code files in a browser. It also provides an optional view of shared REPL sessions.
Teaching games
Gradient Surfers is a running game for ENGSCI 205 revision. Questions and ordering tasks cover machine learning concepts and PyTorch code.
Regularisation Run is a snowboarding game for revising inverse problems and regularisation in ENGSCI 721.
Other projects and experiments are on GitHub.