How we run engineering
You built it, you monitor it, you fix it.
Engineers at Intempt build and maintain a production AI platform serving real customers. This is not a research or prototype environment. You built it, you monitor it, you fix it when it breaks.
What engineering owns
- Production-quality code. End-to-end service ownership - build, deploy, monitor, fix
- RCA after every Critical and Major incident
- Push back on bad specs before picking up a ticket. Collaborate with PMs on requirements first
AI tooling
Engineers run Claude Code sessions with harness and loop engineering for backend services - multi-agent workflows, automated code review, test generation, and deployment checks. Lovable handles frontend prototyping and full builds. Cursor for inline AI-assisted editing. Claude on claude.ai for architecture discussions, debugging sessions, and design reviews. Every engineer has AI agents running alongside them throughout the day. The expectation on output reflects this - you ship more because you work with AI, not because you work longer.
Service ownership
Every service has a clear owner. That owner writes the code, monitors it in production, responds when it breaks, and files the RCA. No throwing code over the wall to an ops team. See our Engineering Principles for the technical standards behind this, and Engineering operations for the automation, alerting, incident, testing, and security layer that owner works inside, graded weekly by the engineering health score.
On-call
On-call is not spread across the whole team. It is a small, curated rotation of engineers who know the platform deeply enough to take a production page alone, find the failing service, and fix or safely mitigate it. Being senior, or owning a product area, does not put you on it - demonstrated platform depth and production access do, and the roster is kept small on purpose. A rotation of people who half-know the system turns every page into an escalation.
Standby and incidents are different, and we compensate them differently. What standby pays, how incident time is credited, and how it works by engagement type all live in On-call Compensation, which is the single source of truth for those numbers.
The real reward is a quiet pager. Standby and incident pay are a floor, not an incentive, and nobody should want more pages. The recognition goes to the work that removes them: the fix that stops a service waking anyone, the noisy alert that is now gone. A quarter the rotation barely fired because someone made it so is the outcome we celebrate, and that person gets the credit for it.
Customer support
Engineers don't respond to support tickets directly. Customer support is AI-first. Confirmed bugs and infrastructure issues land in the engineering backlog as regular tickets. If a customer ticket identifies a production incident, it follows the incident escalation protocol.
Engineering teams coordinate a weekly customer support rotation. The person on rotation triages customer-reported issues and closes the loop directly. Keeps engineers close to real user problems.