← Work
Machine Learning & Perception
Invasive Algae Detection with MiSo and Neural Networks
Classified invasive vs native algae from water chemistry at roughly 0.5% of commercial hardware cost.
A low-cost autonomous system replacing slow manual algae ID. I ran controlled experiments across five 1,000-gallon tubs, logged 10 water properties, and trained neural networks on 50,000+ data points.
A custom Arduino platform — MiSo, a 3D-printed waterproof sensor rig with a KiCad PCB — reached 90.4% on a three-tub task at roughly 200× lower cost than commercial-grade sensing.
