Forward Deployed Engineer, Office of the CEO at Campfire · LumiJobs
mid
$140k–$165k
H1B: no
Campfire Forward Deployed Engineer at Campfire responsible for triaging CEO's AI-generated PRs and making them production-grade. Ensures accounting correctness and reliability while working closely with the CEO and customer priorities.
Responsibilities
Our CEO ships code. With AI coding tools, they turn customer conversations into working prototypes and open PRs directly against our codebase. The ideas are good and the instincts come straight from customers. The code is vibe coded.
Your job is to take those PRs to production.
You'll be the engineer who sits between the CEO's editor and main. You'll figure out what each PR is actually trying to do, then decide what to keep, rewrite or drop. You'll get it through review, tests, migrations and deploy without losing the idea that made it worth building. You'll turn a customer demo that half works into a feature that is correct on a real general ledger with real money in it.
This is a real engineering role with an unusual input stream. You'll work closer to the CEO and to customer priorities than anyone else on the team.
Triage every CEO PR. Read the diff and work out the intent. Decide which it is: ship with fixes, rebuild properly, split into smaller PRs, or push back.
Make it production-grade. Add types, tests and error handling. Fix N+1 queries. Make sure the code follows existing patterns instead of inventing new ones. Generate migrations properly and make them safe to deploy.
Protect the ledger. Nothing posts into a closed period. Every model attribute the code touches actually exists. Approval and audit flows (draft queues, permissions) are respected. Integrations hit sandbox endpoints outside production. Accounting correctness is not negotiable, however good the demo looked.
Keep AI features honest. Every LLM call the CEO wires up gets token tracking and spend-limit enforcement before it ships.
Guard reliability. Errors route through our Sentry conventions, failures report once, and customer-fixable problems surface to customers instead of paging engineers.
Close the loop. When the CEO keeps making the same mistake, fix the root cause. That might mean better scaffolding, clearer conventions in the repo, or tooling that makes the AI generate better code from the start.
Say no well. Sometimes the right answer is "this shouldn't ship." You'll explain why clearly and offer a better path to the same customer outcome.
Qualification
4+ years of shipping production PythonExcellent code review instinctsTaste and judgment about when to patchLow ego and high candorSpeedAccounting fluencyDeep experience with AI coding toolsEarly-stage startup experience
Required
4+ years of shipping production Python. Django, Postgres and Celery strongly preferred; TypeScript/React a plus.
Excellent code review instincts. You can read a 15,000-line AI-generated diff and quickly find the three things that will break in production.
Taste and judgment about when to patch, when to rewrite and when to delete.
Low ego and high candor. You'll push back on the CEO's code routinely and need to be comfortable doing it.
The ability to reconstruct intent from incomplete specs, and to ask the one question that resolves the ambiguity.
Speed. The CEO opens PRs fast, and the queue shouldn't become a graveyard.
