BioNexus is a platform that

BioNexus is a platform that identifies African wildlife from a photo, a video clip, or an audio recording. I built it because most wildlife-ID tools are trained on North American and European species. African biodiversity is some of the richest and most threatened in the world, and field researchers and citizen scientists there often lack a tool tuned to what they actually see.

Under the hood, a fine-tuned Vision Transformer identifies 227 species from images and video frames, reaching 75.9% validation accuracy. An Audio Spectrogram Transformer identifies 72 species from sound, reaching 68.3%. An LLM only explains the result and never makes the identification, so a language model can't override a trained classifier. Uploads go to storage and inference in parallel to keep latency low, and low-confidence results are flagged for human verification.