Stuut AI Stuut AI is hiring a Member of the Technical Staff to build AI agents that automate order-to-cash processes. The role includes building agent harnesses, product APIs, and working with customers to improve agent reliability.
Responsibilities
At Stuut, you'll build AI agents that do real order-to-cash work: collecting overdue invoices, responding to customer emails about invoices and payments, applying cash, and resolving disputes and deductions. You'll build the agents, their harnesses, and the product that makes them reliable in production.
Every customer has different processes, policies, and rules. Our agents adapt to how customers work, not the other way around. They need to use tools, manage long-running tasks, recover from failures, check their own work, and know when to involve a person.
Agent experience is a plus, not a requirement. If you've built reliable workflow or distributed systems, you already understand many of the hard problems: state, retries, and human handoffs. You can learn Pydantic AI and order-to-cash on the job.
Build agents that take receivables work from start to finish: applying payments, chasing overdue invoices and resolving deductions.
Build the agent harness: the tools, structured outputs, guardrails and human approvals that decide when an agent acts, waits or escalates.
Make every agent action traceable, and every write to a customer's ERP safe to retry and easy to review.
Build the product, from APIs and data models to the UI, where finance teams set rules, review decisions and steer agents.
Write evals against real customer outcomes, and turn each failure into a fix in the harness.
Work with customers to learn how AR teams actually work, and turn their edge cases into better agents.
You Might Be a Fit If You...
Have built an agent that calls tools and takes real actions, at work or on your own.
Ask "how would we know if this is wrong?" before you ship, and answer it with evals and traces.
Qualification
Understanding of Pydantic AI framework
Required
Experience building agents that call tools and take real actions
Experience handling long-running agent work including state, retries, and human handoffs
Ability to design for high-stakes environments where wrong actions cost money
