Farm Management System

Executive Summary

We supported a fast-growing agri-tech company in enhancing its farm management platform through advanced data science, digital infrastructure, and user interface design. Over several projects spanning 2–3 years, we developed optimisation models, integrated satellite imagery into analytics workflows, and contributed to decision-support tools that help farmers balance profitability, carbon impact, and biodiversity. Our work ranged from backend model integration via APIs to machine learning pipelines handling complex geospatial data.

Case Study: Building Smarter Tools for Sustainable Farming

Client

A UK-based agtech startup developing a farm management platform used by agricultural professionals to track, plan, and optimize farming operations.

Engagement Overview

Over a span of 2–3 years, our team collaborated on multiple data-driven projects to enhance decision-making capabilities within their platform. We delivered optimization models, machine learning pipelines, and UI/UX improvements—bridging data science with real-world usability for farmers and land managers.

Multi-Objective Farm Optimisation

We built and deployed a decision-support engine that helps users evaluate trade-offs between financial margin, carbon impact, and biodiversity when changing crop or production strategies (e.g. converting from conventional to organic systems).

  • Approach: A multi-objective linear programming model, deployed via a cloud function.
  • Integration: Fully decoupled API (Azure Function) – receives farm data as JSON and returns an optimized strategy.
  • Outcome: Enabled farmers to run realistic "what-if" scenarios and make sustainability-conscious planning decisions with measurable financial insight.

AI for Landscape Classification

We helped implement a machine learning pipeline to classify landscape features from 3D aerial data.

  • Problem: Classify point cloud data (from aerial surveys) into unmanaged vs. managed natural features, like hedges or woodland.
  • Solution: A two-step ML pipeline using K-means for spatial segmentation and a Random Forest classifier for vegetation type detection.
  • Challenges: Handled large TIFF and LAS files from satellite providers, optimised for memory-intensive operations on client VMs.

Feed Optimisation & UI/UX Collaboration

We worked with the client's in-house team to turn a complex animal feed optimization model into a user-friendly tool.

  • Goal: Help farmers adjust feeding strategies based on cost, availability, and animal performance goals.
  • Delivery: Model development + collaborative UI/UX iteration to allow seamless data entry, model execution, and result interpretation.
  • Impact: Created a practical tool that combined business value with sustainability awareness.

Technologies Used

Python (Google OR-Tools, Scikit-learn), GLPK, Azure Functions, QGIS, CloudCompare, custom APIs. Our work integrated into their existing digital infrastructure and complemented their R-based internal workflows.

Collaboration Style

We worked closely with the sustainability team and executive leadership, adapting to evolving needs and constraints while staying focused on delivering value. Our ability to bridge scientific modeling, cloud-native development, and end-user experience design was key to project success.