Principal Software Architect (m/f/d) — HRX Platform

Working language: English · Reports to the Co-Founder & Sr. VP HRX

The short version

You will be in the codebase every day. This is not a governance, portfolio or enterprisearchitecture role. There is no architecture board, no target-state deck, no review committee. You design the domain model, you write the constraints that enforce it, and you ship. We build HRX — an enterprise product that unifies CRM, service management and HR service desk on a single data model. We are small, we are deliberately staying small, and we develop AI-natively: our engineers today work through Claude Code, Cursor and Codex against our own codebase, in our own architecture — with automated review on every check-in. That only works if the architecture is good. AI amplifies whatever you give it — including the mistakes. This role is the seat that decides which one we get. We are not looking for someone to produce architecture documents. We are looking for someone who expresses architecture as artifacts a machine can enforce, and who can tell correct output from merely plausible output at a glance.

What you will own

The domain model. One data model spanning CRM, service management and workforce. This is the hard problem in HRX, and it is yours.
Invariants as enforcement. Our specification layer defines invariants and governance rules. You turn them into state machines, database constraints and contract tests — so that no generated code can violate them silently.
The agent context layer. The architecture files, ADRs and module contracts that our AI tooling loads on every session. When this is right, a new engineer can ask the codebase how something works. When it is wrong, we accumulate debt at machine speed.
Review of shape. You review architectural direction and boundary decisions, not individual lines. Line-level review runs automatically on check-in — you own what those gates enforce. Architecture decision records.
Written, versioned, in the repository, and actually followed.

What we need you to have already done

Designed and operated a multi-tenant B2B SaaS platform in production — through incidents, migrations and scale, not only through the design phase.
Deep transactional data modelling — entities, state transitions, constraints, concurrency. You have strong opinions about nullability, defaults and transaction boundaries, and reasons for them. Warehouse, ETL and analytical modelling is a different discipline and is not what this role needs.
Expressed architectural rules as executable constraints, not only as documentation.
Worked daily with AI coding agents (Claude Code, Cursor, Codex or equivalent) on a substantial codebase — and formed a considered view of where they fail.
Written technical specifications and decision records that other engineers actually used. In this model, the specification is upstream of the code. Writing is a core deliverable, not overhead.
Worked in TypeScript (strict), NestJS, PostgreSQL, Kubernetes. Our stack is fixed: TypeScript/ NestJS/PostgreSQL 16 + Prisma/Next.js/Keycloak/Redis/Kubernetes on Oracle Cloud Infrastructure.

Valueable, not required

Enterprise CRM, ITSM or HR service management domain knowledge
Regulated or GDPR-sensitive environments
OCI specifically — Kubernetes and IaC experience on any major cloud transfers fine
German

This role is probably not for you if

You have not shipped production code in several years. If your last few years were spent advising, evaluating or governing rather than building, this is the wrong seat — and we would rather say so now than at the second interview.
You are excited about AI coding tools but your architecture depth is thinner than your enthusiasm. The failure mode of AI-native engineering is confident, plausible, structurally wrong code produced at volume. Preventing that is the job.
You would want to reconsider the stack in your first quarter. It is settled, and settled for reasons we will happily walk you through.
You want a large team beneath you. There isn’t one, and there won’t be. Leverage here comes from architecture, not headcount.
Your background is enterprise architecture in the TOGAF/ArchiMate sense, or cloud infrastructure engineering. Both are real disciplines. Neither is this one.

How we will assess you

No whiteboard puzzles. After an initial conversation you get a real artifact from our platform: a set of

specification invariants and a defect that AI-generated code actually produced in one of our

evaluation builds.

Express one of these invariants so that it cannot be violated.
Tell us what is wrong with this generated code, and what you would have prompted instead.

Roughly ninety minutes. It is the same work you would be doing here, and it tells us more than any

interview format we have found.

Process

intro call (30 min) → work sample → deep-dive with the Co-Founders (90 min) → offer. We aim to close within three weeks.

What we offer

Genuine architectural authority — this is a founding-level technical seat, not a role inside an existing hierarchy. You report directly to the Co-Founder & Sr. VP HRX.
A small team where the distance between a decision and it being live is measured in days
A platform being built now, on a specification base that already exists — you are not starting from a blank page, and you are not maintaining someone else’s twenty-year-old one either
Start: immediately

We read everything. You will hear back either way.

#J-18808-Ljbffr

CereBree