Technical Support Engineer

Own bugs and production issues end to end: reproduce, diagnose, fix, verify, and confirm the fix held

Drive ambiguous reports to a true root cause across our web apps, APIs, and banking integrations

Orchestrate AI agents through the full diagnostic and repair loop, and catch them when they are wrong

Ship your own fixes as reviewed pull requests, with evidence attached to every ticket

Extend the agent tooling itself. The workflows, MCP servers, and triage agents this team runs were all built in-house, and you will build the next ones

Write the plain-language root cause explanation that goes back to the customer

Find and fix classes of bugs, not just instances. Strengthen alerts, runbooks, and guardrails so the same issue does not come back

Help decide how this function scales as the team grows

2 to 6 years writing production software, ideally somewhere you were close to real customers and real incidents

Strong TypeScript. Comfortable moving between a React frontend, a Node backend, SQL, and a third-party API in the same afternoon

You debug by narrowing, not guessing. You can read logs, traces, and a database and form a hypothesis you then try to disprove

You already work with AI coding agents daily, and you have opinions about where they fail. You can tell when an agent is confidently wrong, and you know what you would never hand one

You have a bias toward proving things. You do not say a fix works, you show it

You can explain a technical failure to someone who has never seen the code, without condescension

You manage yourself. Nobody will assign you work hour to hour, and you will flag a slip before the deadline rather than after

You are comfortable when the process is still being written, because you will be the one writing parts of it

Experience at a fintech, or anywhere regulated, where a bug has consequences beyond a bad user experience

Familiarity with Next.js, tRPC, AWS, Sentry, or a structured logging platform like Axiom or Datadog

Experience with multi-tenant systems where behavior is driven by per-customer configuration

You have built internal tooling, MCP servers, or agent workflows that made other engineers faster

We tell you up front, because ours is unusual and deliberately so.

Intro call, 30 minutes. A conversation with our CTO about the hardest bug you have ever fixed, and how you actually work with AI tools.

A build exercise. We give you a spec for a small service. Expect about four hours, with a week to return it. Use whatever AI agents you like, and send us your session logs. We read your evidence and your written decisions before we read your code.

In person in NYC, 2 hours. You defend what you built, then we introduce bugs into your own codebase and you debug them live on your laptop with your own tools, with two of our support engineers watching.

Founders round, 60 minutes. A video call with our CTO, CEO and CPO, scheduled after the onsite. Not a technical round. That part is settled in step 3 and we will not re-run it.

No algorithm rounds. No whiteboard. We are not going to ask you to invert a binary tree.

A very small team holds production quality for a platform serving financial institutions across the country. You would be joining it early enough to inherit a genuinely good agent workflow and shape what it becomes as the team grows.

The work is also real. Bugs here reach people trying to open a bank account or apply for a loan. If that sounds like pressure rather than tedium, we would like to talk to you.