Obirdability is a KG Labs research initiative, launched in October 2025, to build an open Kyrgyzstan-focused bird identification system that recognises local and migratory species from sound and imagery. Kyrgyzstan sits on major Palearctic flyways and hosts globally significant wetland and mountain habitats, but year-round monitoring is sparse and global birding apps fall short on Central Asian avifauna. The goal is a locally tuned tool that researchers, park rangers, and citizen scientists can rely on — and a record that does not go quiet between October and April when the high-altitude sites become unreachable by road.
The work runs on four threads in parallel.
Datasets. Assembling a regionally representative dataset of bird sounds and images, curated for Kyrgyz and Central Asian species and validated by domain experts. The gap is well known to ornithologists who work the region: training corpora that perform well on European or North American species lose accuracy on Tien Shan and wetland species that are under-represented in global collections.
Machine learning and AI. Training and tuning identification models — combining computer vision and bioacoustic deep learning — adapted to local species, regional dialects of bird song, and field conditions. Reference performance is measured against expert validation, not against external benchmarks built on other regions’ data.
Edge AI devices. Designing low-power edge AI field units that run inference directly on-device at remote sites. Identification happens in the field without depending on connectivity; results are stored locally and relayed when a link is available. The hardware brief is deliberately practical — durable enough for unattended winter deployment, cheap enough to install at multiple sites without a procurement cycle each time.
Open science. Findings, datasets, and tooling are published openly so the work benefits the wider scientific and conservation community in the region.
The project is being developed in close partnership with local and regional ornithologists, whose expertise sets species priorities, validates model outputs, and grounds the work in real fieldcraft. We are also coordinating with protected-area staff and conservation NGOs operating at the target sites.
Planned field installations.
| Site | Designation | Why it matters |
| Chatyr-Kul | Ramsar site, high-altitude lake | Key waterbird stopover on the Palearctic flyway |
| Sary-Chelek Biosphere Reserve | UNESCO biosphere reserve, western Tien Shan | Mountain forest and lake habitat; one of the most significant migratory stopovers in Central Asia |
| Issyk-Kul Lake Biosphere Reserve | Ramsar wetland | Major wintering and migration hub |
| Ozernoe Lake, Chuy Valley | Lowland wetland | Rich passerine and waterfowl activity |
Sites confirmed for Phase 1 deployment; installation timing depends on hardware iteration cycle and field-partner readiness.
Status. Active. The team is currently testing a range of edge AI devices and identification models, and the work has progressed significantly since the project launched. The aim is to publish findings and tooling by summer 2026, ahead of pilot deployments at the four sites above.