JockiBox
JockiBox

We need one more engineer.

We're building AI agents that run inside healthcare facilities - agents that do real work and add real value, not flashy demos.

At JockiBox we're building AI agents that run inside healthcare facilities - agents that do real work and add real value, not flashy demos. It's some of the most interesting agentic AI being built right now, and the team doing it is genuinely exceptional.

The person we're looking for is a software engineer who is already working in agentic AI today - building with it, breaking it, obsessed with where it's going. Someone who solves problems from first principles, thinks sideways when the obvious path is wrong, and takes full ownership - delivering functional, finished products that blow people away.

What you'd be doing

The core is live. You'd be joining a team working to get it from here to something that runs reliably at scale. On-site in Eagle, Idaho.

Build features end to end - From a ticket or idea to a reviewed, deployed, working slice. Frontend, backend, data, and the wiring between them.
Work alongside AI coding agents - Task them, review what they produce, catch errors, and integrate the correct output. Direct and verify capable-but-fallible agents rather than accept their output uncritically.
Own quality - Read diffs line by line and judge whether they're correct. Validate against real behavior - done means it works when exercised.
Help build the playbook - How to decompose work, when to trust an agent, where a human stays in the loop. The patterns are still being defined here.
Design and lead agent teams - Architect multi-agent systems - deciding how agents divide work, hand off to each other, and escalate to a human. You're not just writing agents; you're building the systems that orchestrate them.

What we're looking for

A strong engineering foundation - you've shipped production software and can read code critically.

Already working in agentic AI - building with it, not just reading about it.

Trust-but-verify instinct toward agent output. Low ego about who wrote the code; high standard for whether it's correct.

Self-direction - picks up the highest-value work without waiting for a fully-specified ticket.

Comfort with ambiguity - the patterns here are still being written, and you help write them.

If you love the idea of building something that matters in agentic AI - let's talk.