The Quest
Conservation researchers need to count and identify wildlife across huge volumes of aerial-census imagery. Doing it by hand doesn't scale, and off-the-shelf tools don't fit the specialised workflow.
The Forging
I work as lead developer on SurveyScope in collaboration with WildEye Conservation, building across React, TypeScript, and AWS (Amplify, AppSync, DynamoDB, SQS, ECS) with Python services behind the scenes. Recent focus has been backend optimisation and expanding the platform's workflow capabilities — model-guided annotation, quality-control pipelines, homography generation, and individual animal identification.
The Victory
SurveyScope is live and actively used by conservationists, with ongoing development broadening what researchers can do with their imagery.
Notable Enchantments
- Model-guided annotation— AI assists human annotators on aerial imagery.
- Quality-control pipelines— Structured review flow for census data.
- Homography generation— Aligns overlapping aerial frames so the same ground area can be compared.
- Individual animal identification— Re-identifies specific animals across frames to support tracking and prevent double-counts.
- AWS-native architecture— Amplify + AppSync + DynamoDB + SQS + ECS.