Why join us Brex is the intelligent finance platform that enables companies to spend smarter and move faster in more than 200 markets. By combining global corporate cards and banking with intuitive spend management, bill pay, and travel software, Brex enables founders and finance teams to accelerate operations, gain real-time visibility, and control spend effortlessly. Brex’s AI-native automation and world-class service eliminate manual expense and accounting tasks for customers so they can focus on what matters most. Tens of thousands of the world's best companies run on Brex, including DoorDash, Coinbase, Robinhood, Zoom, Plaid, Reddit, and SeatGeek. Working at Brex allows you to push your limits, challenge the status quo, and collaborate with some of the brightest minds in the industry. We’re committed to building a diverse team and inclusive culture and believe your potential should only be limited by how big you can dream. We make this a reality by empowering you with the tools, resources, and support you need to grow your career. AI at Brex AI Engineering at Brex is redefining how businesses run their finances by building intelligent, autonomous systems directly into the Brex platform. Our teams develop AI agents that don’t just surface insights—they take action, optimizing spend, managing workflows, and making real-time decisions on behalf of our customers. By deeply integrating proprietary financial data with product and platform infrastructure, we’re turning complex financial operations into simple, automated experiences and setting a new standard for how modern finance works. What you’ll do As an AI Engineer on our Finance Agents team, you’ll build Brex’s Agent Workbench: our singular conversational surface where finance professionals work with Brex’s agents through documents, charts, spreadsheets, and other interactive surfaces. You’ll help customers understand financial activity, investigate findings, revise policies, and steer agents doing consequential work. Your first substantial project will be evolving the existing workbench into an extensible platform that can support multiple agent experiences, including bookkeeping and Brex Assistant. You’ll own the extension model, integration protocols, and application architecture that enable other teams to bring their agents and experiences into a coherent workspace. You’ll establish the agentic design system with the technical depth to connect interaction design to platform decisions. Working closely with frontend engineers, agent teams, Product, and Design, you’ll build across application layers as needed and iterate with customers. The scope includes experiences such as policy editing, reporting, backtesting, and interactive data exploration, with priorities shaped by user needs. Where you’ll work This role will be based in our San Francisco office. We are a hybrid environment that combines the energy and connections of being in the office with the benefits and flexibility of working from home. We require three coordinated days in the office per week — Monday, Wednesday, and Thursday. As a perk, we also have up to four weeks per year of fully remote work! Responsibilities * Design and build a reusable agent workbench platform that supports multiple agent personas and enables other teams to integrate their agents and product experiences. * Design and implement extension models, integration protocols, and application architecture that connect agent capabilities with rich, consistent user interactions. * Build and ship customer-facing workflows that combine conversation with documents, charts, spreadsheets, and interactive artifacts, owning UI implementation and the supporting application behavior. * Partner with frontend engineers, agent teams, Product, and Design to turn ambiguous user needs into cohesive experiences, making thoughtful tradeoffs across usability, extensibility, and reliability. * Own features from design through production and iteration: talk directly with customers, investigate failures, and use feedback and observed outcomes to improve how users understand, configure, and steer agents. Requirements * 5+ years of software development experience delivering customer-facing applications * Experience building or contributing to a production grade agent, agent harness, workbench, or agentic product. * Experience in application architecture and integration design, with the ability to translate interaction needs into extensible interfaces, protocols, and clear boundaries between components. * Deep familiarity with agent clients such as Claude, ChatGPT, and OpenCode, grounded in sustained use. * Strong product judgment and ownership in ambiguous environments. You work effectively across engineering, Product, and Design, challenge decisions constructively, and take responsibility for production quality and customer outcomes. Bonus points * Full-stack experience spanning rich frontend interactions, backend services, APIs, and data modeling. * Experience building IDEs, developer tools, real-time collaborative applications, or similarly sophisticated web products. * Experience designing extension systems or reusable application platforms adopted by other teams. * Professional experience shipping LLM or agentic products, including human-in-the-loop workflows, tracing, evaluation, or feedback mechanisms. * Experience with financial software or other consequential workflows where transparency, traceability, and user control matter. Compensation The expected salary range for this role is $216,000 - $240,000 USD.  However, the starting base pay will depend on a number of factors including the candidate’s location, skills, experience, market demands, and internal pay parity. Depending on the position offered, equity and other forms of compensation may be provided as part of a total compensation package. Brex LLC is a wholly owned subsidiary of Capital One, N.A. Please be aware, job-seekers may be at risk of targeting by malicious actors looking for personal data. Brex recruiters will only reach out via LinkedIn or email with a brex.com domain. Any outreach claiming to be from Brex via other sources should be ignored.