Role: GTM Engineer, inbound and outbound systemsFull interview loop

For a software company with a sales team, hiring an engineer to own the systems between a buying signal and the seller who acts on it: how inbound requests and signups are matched and routed, what enrichment is allowed to change, what the CRM holds, and the automated outbound that runs every morning. It assumes the candidate will build on the company's own CRM and data and will own what they build after launch, so there is no round on any one tool and none on writing outbound copy.

The four rounds are the same pipeline seen from four places. Round one is a design conversation about where an inbound request should go, from a brief that leaves out who owns what. Round two hands the candidate a workflow someone else built and its first month of numbers. Round three puts signups into a CRM in code, against a rate limit other tools share and fields reps have already edited. Round four is a system the candidate built, and what the sellers did with it.

The design round comes first, on purpose. It scores what the candidate asks before anyone has told them that a third of inbound requests come from companies already in the CRM, that the enrichment vendor has never been checked, or that the CRM holds reps' own work, and every later round tells them something like that. Matching a person to a company is the work the role turns on and the easiest part of it to describe well, so it is read in conversation in round one and in code in round three, with records that set different traps, and data quality is read in a third round besides: in round two, a model's line built on news matched by company name. Round three reads coding in its first section, where the sync has to survive a resent event and a launch under a rate limit other tools share, data quality in the two after it, and GTM systems design in the last, where the run has to tell someone who is not an engineer what went missing. Commercial judgment is read once in someone else's workflow, where the reply rate is the number on offer, and once in the candidate's own system, where the temptation to claim credit is strongest. Working with sellers is read once live, in a mid-quarter change request from a sales leader, and once in the candidate's past, as a request they turned down. Ownership is read on the morning the candidate inherits round two's workflow, and in what was still running after their own system launched.

Round two is the loop's only read of model-written output: an opening line that congratulated a CTO on funding his company never raised. It is read there as a data problem, because that is what it was. If your GTM engineer will build agents that read what prospects write and act on it - reply to them, update the CRM, book meetings - add the AI Engineer loop's round on untrusted input and tool permissions as a fifth. Its assistant reads support tickets rather than inbound email, and the questions carry over: what the agent may do on its own, and the morning it does too much.

Coding is a nice-to-have, read once, in round three. This loop will pass a candidate who writes plain integration code and asks whose field it is writing over one who writes elegant code and has never had a rep complain. If your GTM engineers spend most of the week in the warehouse and in code rather than in the CRM and a workflow tool, add a general coding round and make Coding a must-have.