A recommendation is a decision made on someone's behalf.
Most of those decisions are guesses in good posture — "customers also bought," a rule someone wrote two years ago, a black box no one can question. When they're wrong, the customer pays in wasted time and the business pays in returns.
We started Specicon on a stubborn belief: a decision you can't explain shouldn't ship. So we built an identity layer that reads behavior instead of storing people, engines that reason before they answer, and a reason string that travels with every match — the same words support, merchandising and the customer all see.
Then we made ourselves prove it. Every rollout runs against a live holdout — because trust is earned in evidence, not adjectives.
Small teams, real ownership, everything written down.
Remote-first, with hubs in the USA and India. We optimize for judgment over process — and we write, because a decision you can't explain shouldn't ship here either.
How the team is organised.
Operators, engineers and scientists who'd rather ship one decision that holds up than ten that mostly work.
…and a growing team across three time zones. See who we're hiring →
Come build the decision layer.
If you can help machines decide well — and explain it — we want to hear from you. Tell us what you'd build.
The work is hard. The rest shouldn't be.
Help machines decide well — and explain themselves.
Whether you build the engines or bring them to customers, there's a seat for people who care about decisions that hold up.
