Substrate Labs

Built in India

AI research and engineering from India.

We build Viveka, models that turn your app’s data into decisions software can act on. Viveka Hoshi, the fourth generation, is in private beta.

Join the waitlist

94.9%
accuracy on 3,000 Open-Jev test decisions
99.7%
right when Hoshi is 90% sure or more, on 2,512 of those decisions
96.5%
of right answers inside its answer set, against a 95% target

Viveka Hoshi

Viveka

The fourth generation of Viveka. Send your app’s data and a question. Get a choice, a score or a yes/no answer, with a probability for every option and a flag when it isn’t sure.

Choose one
Route a ticket, a request or an action to one of the options you define, with a probability for each.
Rate on a scale
Score urgency, sentiment or relevance across levels you set, with a probability for each level.
Yes or no
The probability of yes, for a rule you provide: approve, block or escalate.

An unsure flag you can route on

Every answer comes with the set of options that could be right. When the set holds more than one, Hoshi marks the answer unsure, so your app can act on the clear cases and hand the rest to a person.

Right answer inside the answer set95% target
  • Open-Jev test2.9% marked unsure96.5%
  • Open-Jev validation4.4% marked unsure95.5%
  • Open-Jev, unfamiliar tasks2.4% marked unsure85.7%

The 95% holds on familiar decisions; on unfamiliar task variants it held for 85.7%. See every measurement and the limits.

Join the waitlist

Viveka Hoshi is in private beta, and we’re letting people in a few at a time. Leave your email and we’ll write when there’s a place.

Private betaHow we use your email