Software Engineer, Full Stack (Agent Platform)

Description

About Sapiom

Sapiom is the end-to-end platform that removes barriers to ship and scale agentic products.

We unify what an agent needs to act in the world — compute and sandboxes, memory, identity, spend controls, storage and queues, monitoring — provisioned together as one thing, not handed over as a framework to assemble yourself.

We have assembled a world-class team with deep infrastructure and payments DNA to build the operating system for machines. Our founder ran payments engineering at Shopify for five years and built an autonomous consumer agent company before that. We raised a $35M Series A led by Dragonfly in August, bringing us to $50M total. Accel led our seed; Menlo Ventures and Anthropic are also behind us.

The models are the brain. We're the spine.

About the role

Agents can think. Getting them to act in production is still hard: teams build a good demo in days, then spend months on the cost, reliability, and control problems that follow. The thing breaks when nobody is watching. The bill arrives before the explanation. Nobody can reconstruct what happened. Closing that gap is what Sapiom does.

Agent Platform is the machinery underneath. It places every model call against latency, cost, and quality targets. It reconciles consumption both in the ledger and in the number a customer is reading right now. It scopes the credentials an agent uses so a customer can hand over specific authority instead of all of it.

The role is full-stack because the platform doesn't end at an API. The control a customer feels is the control they can see and change, so the interface is part of the system you own.

This is a senior role, roughly four to seven years of experience. You'd own one system end to end.

What you'd work on

Routing and capacity. Placing every model call against latency, cost, and quality targets in real time, plus the durable scheduling, reservations, and forecasting that keep those decisions sound as demand shifts. One customer cut inference costs by 75% off the back of this. It's what makes running agents economical at all.

Metering and billing correctness. Two charging layers, hold and capture semantics, and 270M+ consumption events that all have to reconcile. Customers are buying the metering itself, so the correctness bar is high, and the same numbers have to read clearly in a dashboard as well as balance in a ledger.

Reliability at sustained scale. We've absorbed a 10–20x increase that kept compounding week over week for months. What reliability should mean at this scale is still an open question, along with the observability, incident practice, and architectural change that would put us ahead of the growth rather than behind it.

The control plane. Credential-scoped permissions, workflow state limits, gateway hardening, and the surfaces where a customer configures and audits that authority. These decisions set how much power a customer can safely hand an agent, and most of them are still unsettled.

Spend controls people can see. Budgets enforced before money moves, and the reporting that makes an agent's spend explicable afterward. This is usually what stands between an enterprise pilot and an enterprise deployment.

You may be a fit if

You've owned a production system end to end and carried the pager for it.

You've worked somewhere being wrong was expensive: billing, payments, ledgers, anything that has to reconcile against a number someone else is also computing.

You've designed for failure up front — timeouts, retries, idempotency, backpressure — and you can name a case where your design didn't hold and what you changed.

You've debugged something in production that nobody had seen before, under time pressure, and written up what happened.

You've run a multi-month project starting from a problem statement rather than a spec, and other engineers built on what you shipped.

You build the interfaces on top of your own systems instead of handing that off.

You use AI tools in your own engineering workflow.

Nice to have: metering, billing, or usage-based pricing at scale; payments or financial systems; API gateways, proxies, or traffic management; LLM inference, model routing, or capacity planning; queues, storage, or observability infrastructure; systems that talk to a lot of third-party APIs; early-stage companies where you set the architecture rather than inherited it.

Applying

A recruiter screen, then a technical screen. If those go well, a three-part loop: an architecture deep dive, a hands-on AI project, and a conversation with our founder.

Everyone on our engineering team carries the title Member of Technical Staff internally.