RustyData

Selected work

Case studies

A snapshot of recent projects. Different domains, same discipline: understand the system first, then model it honestly.

Flagship work

Organisational design · Research platform

SuperScript — reinforcement learning for team allocation

An agent-based model of team formation, plus six optimisers — including hierarchical RL — benchmarked head to head.

  • Agent-based modelling (Mesa)
  • Hierarchical reinforcement learning (Gymnasium + Stable-Baselines3 / PPO)
Read the case study

LLM pipelines · Language analysis

Fair Tales — LLM language analysis of children's books

GPT-4o language analysis over ~200 storybooks — task decomposition, in-context learning and unit-tested prompts.

  • LLM-based NLP (GPT-4o)
  • Task decomposition & prompt development
Read the case study

Data journalism · Social justice

Complicity — follow the money, count the dead

Three timelines of Israel/Palestine on one clock — every figure traceable to a named public dataset.

  • Data engineering & provenance
  • Uncertainty communication
Read the case study

Also delivered

Smaller, sharply-scoped engagements — kept short here on purpose.

Ecology · Enforcement

Agent-based modelling of illegal fishing

A commissioned agent-based model of fleet behaviour for ocean-enforcement analysis.

Read more →

Sanitation · Hardware

Control software for off-grid sanitation

A small, well-scoped software library controlling waterless sanitation hardware.

Read more →

Some client details are anonymised or generalised; full write-ups available on request.