SurveyScope

Aerial wildlife surveys need a path from imagery to population estimates. SurveyScope brings annotation, checks for missed animals, quality control, and analysis into one platform, with AI models helping along the way.

I was already lead developer on SurveyScope at Innoventix Consulting (November 2024–May 2026), in collaboration with WildEye Conservation. When WildEye became an independent company, I moved there full time in June 2026 and continued as Lead Developer. The current stack is TypeScript, Vite, TanStack, React, and Python, with AWS wrapped in Amplify Gen 2. I also built and use my own AI code review and testing tools to check changes before release.

We’ve built SurveyScope into a platform covering model-guided annotation, false-negative review, quality control and user testing, homography creation, chain linking, and Jolly II results. I’m proud of how far it’s come. These days I spend a lot of time on performance and usability, which sometimes means rebuilding functionality I’ve already built.

• Model-guided annotation— AI models help with annotating aerial survey imagery.
• False-negative review— A workflow to find animals the models missed.
• Quality control & user testing— Quality control workflows and user testing are part of the platform’s development.
• Automatic & manual homography creation— Aligning overlapping images for comparison.
• Chain linking— Identifying the same animal across overlapping images to avoid counting it twice.
• Jolly II results— Result generation and analysis for population estimates.