Engineering at Clera

Teaching an agent to judge talent.

Clera reads someone's background, works out which of 2,000+ open roles at 900+ startups they'd actually get, and makes the warm intro to the founder. A small team builds it with coding agents doing most of the typing, and we merge about a hundred pull requests a week.

A teammate at a desk in the San Francisco house, the balcony door open beside her
A desk in the San Francisco house

How we build

The product, the problems, and the stack.

  1. What we're building

    Three parts: a matching engine that ranks people against roles, an agent that talks to candidates over WhatsApp, iMessage, LinkedIn, Slack and web chat, and the tools our talent team works in all day.

  2. The problems

    Is this person actually a fit for this role, before anyone has interviewed them? When should an agent that's talking to thousands of candidates stop and hand over to a human? Did that prompt change get more people hired, or just more replies?

    None of these have a textbook answer. You get real outcomes back, interviews and hires, and work it out from there.

  3. How you'd work

    You own a slice of the product end to end. An agent writes the first draft of most of the code. You decide what's worth building, point it in the right direction, read the diff and ship it.

    When it breaks, it's yours to fix, whoever typed it.

  4. The stack

    Language
    TypeScript, everywhere
    Web
    Next.js
    Data
    Postgres with Drizzle, Typesense for search
    Jobs
    Trigger.dev
    Models
    Vercel AI SDK, traced in Langfuse
    Coding
    Claude Code

    Well-documented, slightly boring tools on purpose, so the time goes on the problems above.

The team on a headland above the Golden Gate Bridge
The back of a Clera t-shirt reading AI took my job (search)
Seven of the team in a circle, photographed from below against a clear sky

A dozen of us, in a house in San Francisco.

Open source

The tooling we built to run our own infrastructure is public on github.com/getclera. Every repo runs in production.

If you'd rather review a hundred diffs than write ten, talk to us.