Introducing Audtheia, open-source AI for biodiversity monitoring
Hello everyone, and thank you for having me!
I’m Andy Portalatin, a marine biologist. My master’s research used deep learning to classify marine sponges, and that work grew into Audtheia, an open-source environmental intelligence platform that helps scientists and conservationists monitor ecosystems.
Audtheia comes in two approaches.
Audtheia V1 is a cloud-based agentic workflow. It detects species in live video, then nine specialized AI agents analyze each observation against open scientific sources like GBIF, iNaturalist and NOAA, producing 72 data points per sighting and daily PDF reports.
Audtheia V2 runs fully offline. Raspberry Pi 5 field stations with Hailo AI detect species on camera and audio, and a desktop hub verifies every detection, analyzes patterns and generates reports, with no internet needed at runtime. Every data point is labeled by its source, so measured and AI-inferred data never get mixed up. It’s built for long-term monitoring and Marine Protected Area work, where connectivity is rare and data integrity matters.
Both are MIT-licensed and citable through Zenodo.
V1 · github.com/AudtheiaOfficial/audtheia-environmental-monitoring…