Sr. Software Engineer - AI-First

We're not looking for someone who uses AI tools. We're looking for someone who has moved past that — into building the systems that make AI output reliable, consistent, and self-improving.

In practice: you invest as much time improving your development system as you do shipping features. When AI output isn't right, you don't manually fix it and move on — you encode the correction as a rule, a hook, or a reusable skill so that category of problem doesn't happen again.

The Sr. Software Engineer has a sufficient understanding of the product and purpose of their team and is meaningfully involved in discussions related to feature implementation, technical design, risk, and delivery estimates.  They are capable of identifying dependencies ahead or preparatory work is required to keep the team unblocked.  They are able to context switch between concurrent streams of work, e.g. major pivots in prioritization or emergent live issues, as needed.  They are able to operate within a large codebase and understand its structure, overall system architecture, as well as its connection to external processes (e.g. CI/CD, automated testing, non production environments, linters, etc.).  They can fluently discuss the current health of the current base particularly as it pertains to trade offs between technical debt vs new feature development as well as dependencies that may influence time estimates. 

You'll work across a multi-service architecture spanning a Django/Python backend, React frontends, and a shared component library. You'll own features end-to-end — from database schema to UI component.

What you'll do:
- You maintain living context files (CLAUDE.md or equivalent) encoding your project standards and learned preferences — and update them every time you correct AI output.
- You design verification systems — hooks, CI gates, acceptance criteria — so you don't need to review every AI output line by line. You build the system that verifies; you don't become the system.
- You treat every repeated manual correction as a signal that your system needs improvement. Bug fixes become rules. Patterns become reusable skills. Preferences become persistent context.
- You can run multiple agents on concurrent workstreams and coordinate their output without losing coherence.
- You catch what agents miss — logic errors that pass tests, subtle architectural drift, hallucinated APIs — and you improve the system rather than just patching the output.
- You allocate real time to system improvement, not just feature delivery. You understand the compound effect.