Technology
Hypha identifies content, ads, brands, platforms, and the people actually watching, frame by frame and across screens.
Every new source of observed behavior improves our models and expands the data we can sell.
Scale from today's production footprint to 5,500 consented homes.
How we get paid
Annual panel data licenses.
What if Hypha models ran on the TV itself? Our models already run on NVIDIA Jetson chipsets, so the path is real: an OEM ships our models inside their hardware, and we scale without ever putting a Hypha box in a home.
How we get paid
License the models and the rights to the data they produce.
What if Hypha ran inside the systems that encode and deliver video? Every stream, every household, every screen, measured at national scale, with nothing installed at all.
How we get paid
Population-level data licenses.
Maybe we’ll stop when we’re capturing consented, anonymized data from a billion humans worldwide.
Covering content identification, ad and brand detection, and viewer attribution.
A proprietary training set that functions as pre-training data nobody else has.
Plus the operational system to recruit, support, and retain it.
Customers are building the dataset into their own models and products today.
U.S. Patent Nos. 10,932,002; 11,570,513; 12,238,372.