Senior Software Engineer

You own what must be true.

We are hiring a Senior Software Engineer, and the job is not quite the one most of us trained for. More of the building, the known work, the patterns we have done a hundred times, is now handled by AI agents. They write the code, and increasingly the spec. What they cannot do is decide what must be true, and that is where you come in.

We were the engine for decades, writing every line by hand. The engine is now the agent, which frees you to be the engineer: the one who knows what must be true and can vouch for it. You will still build, run and own real services, in NestJS and TypeScript on Postgres and AWS, with a "you build it, you run it" mindset. The part we hire for is the judgement around all of it.

• Name what must be trueDecide the commitments for your area, the invariants that can never quietly break: balances reconcile, the audit trail holds, customer harm stays bounded. Hold the domain and system knowledge an agent has no way to know, and convey it so it can act on it.
• Direct the buildTurn intent into safe, reviewed, working change. The agents do more of the typing now, so you set the direction, make the design calls, and own the result.
• Make it checkableBuild so correctness can be proven rather than hoped for: strong tests, observability with Sentry and CloudWatch, and guardrails that catch a bad change before it reaches an investor.
• Run what you buildOwn deployment, CI/CD and infrastructure as code with Terraform, and the day-to-day health of your services.
• Improve how we buildMake the whole team faster and safer through better patterns, tooling, prompts and checks. Leave the way we build better than you found it.
• Answer for what goes liveStay accountable for what reaches real investors, and write the knowledge down so the next person does not inherit your assumptions as bugs.

• Deep knowing as your edge: the domain, the product and the system, and the judgement to decide what must be true and what must never be allowed to be false.
• Several years building and running production services, enough to know what good design looks like and why it matters.
• Expert with TypeScript and Node.js (we use NestJS) and a relational database like Postgres, and comfortable across APIs, AWS and infrastructure as code (Terraform).
• Fluent at directing AI tools like Claude Code and Cursor, and clear-eyed with them: when an agent writes the spec, you can still catch the fault it quietly baked in.
• A reliability instinct: monitoring, observability, and making correctness checkable rather than assumed.
• A habit of writing things down, because knowledge that stays in your head becomes someone else's bug.
• Clear communication for a hybrid team spread across several timezones.

If you don't tick every box but you work the way we have described, write to us anyway.

The boundary keeps moving.

The split is between known work and judgement work. Known work is the set of patterns we understand well enough to hand off, and the agents keep getting better at it. Judgement work is deciding what must be true and owning it: which commitments matter, what happens when they conflict, and what we are willing to answer for.

Knowing is the scarce input. Each step, from knowing what to build, to the spec, to the code, loses a little, and agents are now good at the last two. Writing it down was never the hard part; knowing what to write is. An agent can learn what is true about the system, but only you can decide what we will not permit to be false.

And the boundary keeps moving. What feels like judgement today becomes routine tomorrow as the models improve. We want someone who pins themselves to what doesn't move, rather than to the tasks they do by hand today. This is what a craft does as it learns, and we would rather you grow with it.

We review applications on a rolling basis, so earlier is better. Send a short note on what you have built and where you have owned what must be true, with a link to your work, and we will take it from there.