Founding Integrations Engineer

We're building at the frontier of what applied AI can do inside real work — the operational core of an industry that moves the physical economy. One of the largest on earth, essential, and almost entirely untouched by modern AI. It still runs on people, spreadsheets, and software written before the internet.

We went vertical first on purpose. Vertical is where the hard problems live: messy inputs, real consequences for being wrong, decades of institutional knowledge nobody wrote down. Anything that works here has been tested against reality in a way horizontal tooling never is.

And it doesn't stay vertical. The systems we're building — how work gets decomposed, verified, corrected, and learned from — aren't specific to one industry. They're specific to work. The vertical is the proving ground. The reapplication is the company.

We're deliberately quiet about which industry until we talk. What we'll say now: it's enormous, it's overlooked, and the incumbents aren't coming. You'll get the full picture on the first call.

• A cap table most early-stage companies would envy. Raised at the top quartile of seed-stage rounds by size — 36+ months of runway — from the seed investors who were early in Palantir, Databricks, Anduril, GitLab, Retool, Lyft, Square, DoorDash, Superhuman, and Ironclad.

• Real traction, right now. Live in production with design partners, and the data flywheel is already turning. Demand isn't the bottleneck — execution is. The market is tens of thousands of enterprises, each worth seven figures a year.

• Founded by multi-time exited operators. The founding team has built and sold multiple software companies and spent years up close with dozens more. You're joining people who know how this is actually done — not learning it alongside them.

An AI system that can only read is a smarter dashboard. The moment it earns the right to act — to turn a decision into a transaction, and write that transaction into the customer's system of record — it stops advising and starts working. That moment is the whole game. You own it.

You're the founding engineer for integrations. You own how we connect to, and write into, the operational platforms our customers run their businesses on. You start with CargoWise, the platform behind roughly 80% of our market, and grow into owning the entire integration surface: Magaya, Descartes, and whatever comes next.

The CTO owns the canonical platform you build on — the data layer, the connector architecture, the schemas. You own everything that lets that platform reach into the real world and change it.

You're being handed the keys to modify customers' systems of record — the systems their entire operation depends on. A read that's wrong is a bug you fix later. A write that's wrong is a corrupted shipment, a bad customs filing, a broken invoice, inside the platform a customer runs their business on.

The hard part isn't moving data. It's bidirectional write-back done with the safety the stakes demand: idempotency and replay, not triggering the customer's own workflows by accident, respecting each platform's validation rules, reconciling when our view and theirs disagree. If that responsibility reads as a warning, this isn't your role. If it reads as the most interesting problem on the board, keep going.

The bar is integration depth and judgment under real stakes. Here the domain isn't optional — a serious integration engineer already understands this industry, because you can't move shipment, consol, customs, and rating data correctly without knowing what it means. That's the inverse of most AI roles: the platform fluency is the domain fluency, and neither can be faked.

• You've shipped bidirectional write-back into a production TMS or comparable system of record. (Hard requirement.) Not read-only feeds, not one-directional pulls. You can walk through specific things that went wrong in production — a duplicated transaction, a workflow you triggered by accident, a reconciliation conflict — and what you did about it.

• Deep, hands-on CargoWise / eAdaptor experience. The SOAP/XML surface, eAdaptor Next, certificate-based auth, the warts the docs don't mention.

• You see the common model beneath different platforms. Shown two systems, you instinctively separate the shared business logic from the platform-specific access layer. That's what makes you a category owner rather than a single-platform specialist.

• You build inside someone else's architecture. Asked how you'd bring CargoWise into our canonical connector pattern, you start by mapping it in — not by explaining why CargoWise has to be special.

Python 3.12 · FastAPI · Celery · PostgreSQL · Redis · AWS. You'll build write-back as connectors behind our BaseConnector abstraction — bringing deep platform knowledge into our patterns, not bolting a platform-shaped architecture onto the side. You don't need to be a Python expert on day one, but the eAdaptor world is SOAP/XML and often .NET-flavored, and this isn't — so you should be motivated to build in our stack rather than around it.

Know someone exceptional? Introduce us. If we hire them into a full-time role, we'll send you $5,000.