AI Agent Engineer | Trace

Ideally three days a week in the office, but we’re flexible.

About Trace

Trace is building AI agents that do freight operations work: handling quote requests, coordinating shipments, following up, and working across the systems logistics teams already use.

Learn more at trace.io.

We’re looking for someone who can take an agent from an idea to a system people can depend on.

You’ll work directly with Noam, our founding engineer, and Lucas, our CTO. You’ll help shape both the agents our customers use and the internal agents that help us build and run Trace.

How we work

We use Claude Code and Codex extensively , supported by tool integrations, development harnesses, and automated validation. We expect you to be comfortable working this way and to help us improve it.

The engineering bar is high. You need to understand what an agent is doing, why it failed, and how to demonstrate that a change makes it better.

What you’ll do

• Build customer-facing and internal agents that complete workflows across tools and systems.
• Design agent harnesses: tool interfaces, context management, memory, orchestration, and execution controls.
• Connect agents to APIs, business applications, and the information they need to act.
• Build evaluations, regression tests, and realistic test environments; inspect execution traces and verify actual outcomes.
• Handle ambiguous instructions, failed tool calls, retries, permissions, and knowing when to involve a person.
• Improve reliability, response time, and cost as agents take on more work.
• Use automated coding tools to accelerate development while reviewing and validating the resulting software.

You’ll fit here if

• You’ve built agents that use tools and carry out multistep tasks, and you can show us how they behave.
• You understand the software behind them well enough to debug and change it.
• You can explain how you manage context, evaluate behaviour, and recover from failures.
• You test difficult cases, including whether retries can duplicate actions or an agent can report success without completing the task.
• You choose frameworks and abstractions because they solve a problem, and can build directly when that makes more sense.
• You experiment quickly, question assumptions, and care about what happens after the demo.
• You’re independent, attentive to product quality, and comfortable taking ownership in a small team.

We care about your ability to build, your judgment, and how quickly you learn.