SOSCHI — UN Climate-Health Platform & climatehealth R Package

Overview

SOSCHI (Standards for Official Statistics on Climate-Health Interactions) is a £5m Wellcome Trust-funded programme hosted on the UN Global Platform. As Technical Lead I own the platform engineering and the open-source statistical tooling that makes the SOSCHI methods usable by national statistical offices.

In March 2026 the methods were endorsed by the United Nations Statistical Commission at its 57th session and included in the Global Set of Environment and Climate Change Statistics.

My Role

Technical Lead — platform engineering across services, APIs, deployment, monitoring and maintenance.
Led the team that built the climatehealth R package and its Plumber API: I set the design and engineering standards and matrix-managed the developers who wrote the functions and modules, coding where the team was stretched. Published on CRAN; past 2,000 downloads since April 2026.
Indicator modules delivered as R Shiny applications with partner analysts, including one operational at the National Institute of Statistics of Rwanda.
Drafted the hosting contract that took the platform into production and saw the security assurance through to sign-off; drafted the statement of work for an external supplier and reviewed their deliverables against it.
Manage the GitHub organisation, RAP standards, code review, pre-commit hooks, unit testing and pull-request templates.
Line manage two data scientists and direct technical contributors across five countries.

The climatehealth Package

Six climate-health indicators: temperature-related mortality, wildfire smoke exposure, air pollution, suicides related to extreme heat, malaria, and diarrhoeal disease.

Methods: distributed lag non-linear models (DLNM), quasi-Poisson time-series regression, case-crossover analysis, Bayesian spatio-temporal models using INLA, and multivariate meta-analysis for sub-national estimates.

A single tested implementation serves three surfaces — the Flask web platform, an R Shiny dashboard, and command-line users — through the Plumber API.

Platform Engineering

Flask web application exposing the climatehealth workflows through API endpoints.
Data ingestion (Copernicus ERA5, OpenAQ), search and indicator pipelines.
Dockerised services on AWS orchestrated with Kubernetes; CI in GitHub Actions, CD into the secure environment; semantic versioning; instrumentation for latency, error budgets and drift.
PySpark/Databricks pipelines with MLflow tracking, automated evaluation gates and data quality contracts.
A parallel Rust reimplementation (climatehealth-rs) of core algorithms for performance benchmarking, and a JavaScript data explorer for interactive visualisation.

BREATHE

ONS technical lead for BREATHE (2025–2028), a Wellcome-funded programme led by the University of Bristol across eight institutions in five countries, running causal attribution over a 100-million-person cohort. I own the analytical design and build of the ONS contribution.