How We Build at Roe

We are a small team building AI agents that investigate financial crime for risk and compliance teams. Our customers include companies like eBay, Affirm, and Crossmint. The work is technically hard, the stakes are real, and we are hiring engineers who want to build at the frontier of what is possible with AI.

Before you apply, you should know how we think about engineering. Not our tech stack. Not our interview process. How we actually work.

Product-driven engineering

Every engineer at Roe is a product thinker. There is no product team handing specs to engineering. Engineers understand the why behind what we build and actively shape what is next.

Care about the problem, not just the code. Talk to customers. Watch them use what you built. Have opinions about what we should build next, and those opinions will matter. If you want to be handed tickets and left alone, this is not the right place.

Customer-first mindset

Every engineer at Roe is customer-facing. You join customer calls. You read their Slack messages. You hear their frustrations directly, not filtered through three layers of account management.

The best product decisions come from engineers who deeply understand the problem space. The only way to deeply understand it is to talk to the people who live in it.

Extreme ownership

Engineers own their areas end-to-end: design, implementation, testing, deployment, on-call, long-term reliability. Harder than working in a larger org. Also more satisfying.

Ownership means caring about outcomes, not outputs. You are not done when the PR merges. You are done when the thing works in production and customers are getting value from it.

Move fast, validate faster

If you have a good idea, prototype it. Today. Demo it tomorrow. No approval, no design review, no committee. The best ideas often come from engineers who see something in the data. We want those ideas built and validated before anyone has time to talk themselves out of them.

Use AI aggressively

We build AI products and we use AI to build them. Cursor, Vercel, Claude Code. We expect engineers to use them extensively. The engineers who figure out how to work effectively with these tools will build 50× what they could have two years ago.

Reviews over raw code

When AI can generate code quickly, code review becomes more important, not less. The bottleneck shifts from writing to ensuring correctness and architectural integrity. We weight your ability to review code critically over your ability to produce it quickly.

Why now

AI agents are becoming capable of real work: production systems handling complex, high-stakes tasks autonomously. The next few years will determine which companies build reliable, trustworthy AI systems and which do not. Staying ahead means hiring engineers who are energized by this moment.