d) - Agent SDK

voize

1d•Hybrid

About The Position

At voize, we believe the greatest gift to frontline workers is time - time to care, connect, and be present. Today, that time is lost to busywork and complex systems that pull them away from what matters most: people. Our vision is to change that by building AI companions that seamlessly take over digital workflows. We don't replace humans with technology - we amplify their impact. Our mission is backed with a $50M Series A funding led by Balderton Capital, with support from HV Capital, Y Combinator and other leading VCs. Today, 2,000+ facilities trust voize, and over 200,000 users rely on our AI companion to ease their daily workload. As a dynamic team, we combine first-in-class technology with meaningful social impact. And now, we're looking for you to join us on this mission! As Senior Backend Engineer on the team behind our Agent SDK, you build and maintain the Kotlin foundation that our AI agents move onto. Agents have grown into a product surface across verticals and markets, and the layer underneath them needs a permanent owner. Your work makes the cost of the second and third agents visibly less than the first, so feature teams can ship at their own pace.

Requirements

Several years of production backend experience on the JVM — Kotlin, Java, Scala or similar.
Readiness to work in Kotlin.
Experience designing APIs that other engineers build against.
Experience with an SDK or an internal platform where the contract is the product.
Experience building and maintaining integrations, and keeping them stable when the system on the other side changed.
Testing and observability are part of how you build.
Drawn to AI agent systems, and to the orchestration and evaluation work underneath them.

Responsibilities

Build and maintain the SDK as an internal product.
Shape the runtime and the client-facing contracts that consuming teams build on.
Help set the roadmap for the SDK.
Build the identity layer that agents run on, so customer-facing agents can write with a real user identity.
Design and ship the auth plumbing.
Make every agent observable from the day it launches.
Wire telemetry and error reporting into the SDK by default.
Integrate inference and anonymization so the SDK enforces them for every consumer.
Build the test kit that the SDK still needs.
Create architecture tests that protect the runtime internals.
Build the path that runs every SDK agent against our ML team's evaluation harness before a model change ships.
Absorb ecosystem change on behalf of consuming teams.
Own Koog version bumps and format compatibility.
Set the shared limits on tool output and token usage.