Senior QA Engineer for AI Platform

*Note, you will only be considered if you provide the requested Loom We run an AI-native revenue platform for Property Tech + B2B clients: a Next.js client portal, a Supabase/Postgres data backbone, Python data pipelines, and workflow automations, built and operated by a small senior team. Over the next quarter we are continue to productize and are releasing a number of features into beta with real users, and we need a dedicated engineer whose job is to find bugs before our users do, then fix them. This is a QA and engineering hybrid. You own quality for each release, and you ship your own fixes. What you will own - Pre-release verification of every beta feature. You drive the feature against live staging data and prove it runs, never just that the test suite passes. - Our systematic bug-hunt process. It already exists: multi-dimension code review with an adversarial verification step and a severity ledger. You run it, extend it, and act on what it finds. This will serve as your co-pilot so you will need to continue to optimize and own - Regression tests at the seams where our bugs actually live: service boundaries, database migrations, async UI paths. - Shipping fixes end to end through our PR review process, in TypeScript and Python. The bugs we actually fight (recent, real, anonymized) - Silent failures: a scheduled job that reports green while a gate condition quietly stalls every record it touches, forever. - Migration drift: a merged database migration that never reached production, so the feature shipped dead while the deploy health check matched perfectly. - Boundary mismatches: two ID spaces compared without translation, so a cache guard could never match and every request silently paid the cold path. - Tests that lie: unit tests that passed because they encoded the same wrong assumption as the code they tested. - Unguarded awaits that leave a UI stuck in a permanent "sending" state after a transport error. If you read that list and started matching it to bugs you caught yourself, you are who we are looking for. Must-haves - 5+ years of professional software engineering. You built and debugged production systems before AI assistants existed, and it shows in how you work. - Deep PostgreSQL. You can prove what is actually deployed (constraint definitions, information_schema, function bodies), reason about migrations and row-level security, and debug scheduled jobs. - TypeScript (Next.js/React) and Python, both at a level where you fix what you find. - A verification habit. You check behavior against live data at the real seam. A green test suite is a starting point, never the proof. - Clear written English. Your defect reports carry reproduction steps, evidence, and blast radius: did this fire in production, or is it latent? On AI tools (read this part carefully) We use Claude Code heavily and expect you to be fluent with it or a similar agentic coding tool. It must amplify your engineering, never replace it. You read diffs yourself, form your own hypotheses, and use AI to move faster on work you already understand. If your instinct when something breaks is to regenerate the code until it passes, this is not your role. Our paid trial task is designed to detect exactly this. Logistics - At least 4 hours of daily overlap with US Pacific time - Paid trial task for finalists: a real bug hunt on a scoped slice of our codebase To apply, answer the screening questions below. Start your application with the word "latent" so we know you read this far. Reply with A loom that demonstrates similar work product. You will only be considered if a Loom is provided 1. Describe a silent production failure you personally caught. How did you prove whether it had actually fired versus sitting latent? 2. A database migration merged and CI is green. How do you verify it actually applied to production? 3. Tell us about a bug where the tests passed and were wrong. What did you change about the tests afterward? 4. How do you use AI coding tools day to day, and in which situations do you not trust them? 5. What is your experience with Supabase or raw PostgreSQL at production scale? Be specific about schemas, migrations, and scheduled jobs.