Skip to main content

Cultivo Land

Cultivo is a public benefit corporation accelerating investment into nature regeneration. They partner with NGOs and land stewards to co-design, fund, and monitor land regeneration projects to remove carbon, improve water retention and boost biodiversity.

Visit website
  • Head of Technology
  • Engineering Lead
  • Full-stack Developer
  • Solutions Architect

A new land assessment platform

When I joined Cultivo they had geospatial analysis tools to determine the natural capital potential of a land asset, but they had an entirely manual process for creating assessments PDF documents.

I architected and constructed a platform to present natural capital assessments as secure, interactive web pages. I also automated the manual multi-step process of running the geospatial analyses with an automated workflow, optimising it from multiple days of manual work to one hour of automation.

An assessment showing carbon sequestration potential.
An assessment page showing a map and summary natural capital potential.

Architecture

I'll make an infrastructure diagram some day. For now I'll just throw the tech stack in here. Click the box to see some fun physics.

Next.js
Vercel
TypeScript
GCP Cloud Run
GCP Workflows
Hasura
PostgreSQL
Google Earth Engine
Permit.io
Sanity CMS
Pulumi
GitHub Actions
Storybook
SCSS
Material UI
framer-motion
styled-components
i18next
GraphQL

ML to the rescue

On the path to fully automating the initial land assessment, there was one crucial step which had always required an environmental scientist to intervene - to decide on the most appropriate land restoration activity for the area of interest.

During a 3 day hackathon I created a machine learning model to recommend a restoration activity for a site based on an initial set of geospatial analyses covering aridity, biome, temperature, land-cover and degradation. The model had > 97% accuracy, categorising to 5 restoration activities.

Pipeline map

The Cultivo platform pipeline showing a selected land regeneration project.

As the scope of the company's activities grew it became necessary to visualise the entire pipeline of a land regeneration projects and to support filtering, searching and data aggregation. I implemented this map-based interface using Mapbox, which does a whole lot more than the screenshot can show.