AI Engineer, Workbench

The role

This role sits on the Finance Agents team and centers on the workbench that lets finance professionals converse with agents while reviewing documents, charts, and spreadsheets. The engineer will shape the platform so multiple agent experiences, such as bookkeeping and the Brex Assistant, can share one architecture. The work spans interaction design, integration protocols, and frontend implementation, with close collaboration across engineering, product, and design.

What you would do

• Create a reusable workbench platform that hosts several agent personas and lets other teams add their own agents and experiences
• Define extension mechanisms, integration interfaces, and overall application structure linking agent abilities to polished interactions
• Ship customer-facing flows that mix chat with documents, graphs, spreadsheets, and interactive outputs, including the UI and supporting logic
• Work with frontend engineers, agent teams, product managers, and designers to resolve vague user needs into unified experiences and weigh usability against extensibility and reliability
• Take features from design through production and beyond, speaking with customers, diagnosing failures, and using feedback to improve how people configure and direct agents

What they asked for

• Five or more years building customer-facing software
• Hands-on work on a production agent, agent harness, workbench, or agentic product
• Skill in application architecture and integration design, turning interaction needs into flexible interfaces and well-defined component boundaries
• Working knowledge of major AI assistant clients, gained through regular personal use
• Sound product instincts and comfort owning outcomes when requirements are unclear
• Collaborates well across engineering, product, and design, pushes back constructively, and stands behind production quality and customer results

Nice to have

• Full-stack background covering rich frontend work, backend services, APIs, and data modeling
• Prior work on code editors, developer tooling, or live multi-user web apps
• Experience creating extension frameworks or shared application platforms used across teams
• Shipped LLM or agent products that include human review loops, tracing, evaluation, or feedback capture
• Background in finance software or other high-stakes domains where auditability and user control are critical