Forager

After a rainy week in March, the engineering courtyard filled with at least six different species of fungi. Forager was built to answer the obvious question: which ones are edible?

The app uses a CoreML model trained on a dataset of ~1,400 labelled photos taken on campus and from public domain collections. Point your phone at a mushroom, get a species name, a confidence score, and a clear EDIBLE / AVOID indicator. The model runs entirely on-device — no network needed.

Training and labelling was done collaboratively across a weekend. The model currently identifies 22 species with 84% top-1 accuracy on the test split. A Figma-designed UI wraps it with a field-guide aesthetic.