Neblina Labs wins a WILDLABS Award

Project: Open-Source Bioacoustic Classifiers for Western Amazonian Restoration, Built with Shuar Women

Neblina Labs has been awarded a WILDLABS Award for a project building open-source bioacoustic species classifiers for the western Amazonian foothills of Morona Santiago, Ecuador, developed with Shuar women who manage traditional Aja agroforestry systems.

What the award is.

The WILDLABS Awards, run by WILDLABS with support from Arm, fund conservation technology projects that use Arm-based hardware and share their results openly. WILDLABS is the largest global community for conservation technology.

The problem.

Bioacoustic foundation models cannot identify most species in this region. Not because most species are rare, but because training data from here never reached the institutions building the models. Restoration finance requires biodiversity evidence at a rigour and scale no existing tool can produce in these forests. What cannot be measured cannot be funded.

What we will do.

• Deploy 10–12 AudioMoth recorders across a restoration gradient: Ajas of different ages, cattle pasture, and primary forest.

• Train species classifiers with Shuar women annotators who already identify dozens of bird and frog species as part of daily practice. An active-learning pipeline (agile modeling) cuts expert annotation effort roughly 25-fold; clips are distributed for annotation over WhatsApp.

• Run the entire pipeline: embeddings, vector search, linear-probe classifiers, on a single Jetson Orin Nano. No cloud dependency, which matters where internet and grid power are unreliable.

• Release trained classifiers, pipeline code, deployment protocols, and reference samples openly. Annotation data carrying traditional ecological knowledge—Shuar names, ecological associations, habitat notes—stays under community governance.

Why it matters.

Ajas are polycultural plots integrating up to 70 plant species while sustaining soils, water, and forest structure. The techniques work, and we believe they can restore land cleared for monoculture and cattle. Making that restoration measurable is what connects it to conservation finance.

The Team

• Nicolás Schuldt Durán (Neblina Labs), project lead.

• Enriqueta Apik, traditional ecological knowledge lead and primary annotator.

• Fernando Tsukanka Huambutzereque Chinkim, community governance.

• Camilla Søtorp Albán (LSE), human dimensions.

• Elias Viteri-Basso (USFQ), field biology.

• Gabriel Nunez (2050 Advisors), community-led monitoring.

• Pedro Galindo (Fundación Jocotoco), bioacoustics advisor.

• Ana María Durán Calisto (Yale), advisor on Indigenous territorial governance, intercultural research design, and ethics.