Lead AI Engineer - 11510

Mexico·Mexico CityRemoteMid-Senior Levellead

OtherLead Ai Engineer

0 views0 saves0 applied

Apply Now

Quick Summary

Overview

Technical Tools

anthropicpythontypescript

Coupa makes margins multiply through its community-generated AI and industry-leading total spend management platform for businesses large and small. Coupa AI is informed by trillions of dollars of direct and indirect spend data across a global network of 10M+ buyers and suppliers. We empower you with the ability to predict, prescribe, and automate smarter, more profitable business decisions to improve operating margins.

Why join Coupa?

🔹 Pioneering Technology: At Coupa, we're at the forefront of innovation, leveraging the latest technology to empower our customers with greater efficiency and visibility in their spend.

🔹 Collaborative Culture: We value collaboration and teamwork, and our culture is driven by transparency, openness, and a shared commitment to excellence.

🔹 Global Impact: Join a company where your work has a global, measurable impact on our clients, the business, and each other.

Learn more on Life at Coupa blog and hear from our employees about their experiences working at Coupa.

We're looking for an Applied AI Engineer to build the harness that supplier agents run inside. The harness is not one thing. It's the eval pipeline, the context management layer, the sub-agent orchestration patterns, the document parsing that turns supplier uploads into usable agent context, the CLI tool layer around Coupa APIs, and the skill framework. You'll own pieces of several of these and rotate through the ones that need the most work.

This is a role for a strong AI native engineer who wants to go deep on agentic systems and isn't waiting for permission to experiment. You'll ship often, iterate on real user feedback, and level up fast.

Build and extend the eval pipeline: how we measure task completion, catch regressions, and turn production failures into fixed behaviors.

Work on context management: what the agent sees, when, and how we keep long-horizon tasks coherent without burning tokens.

Design sub-agent patterns: when to fan out, how to compose specialized agents cleanly, and how to keep the parent agent in control of the outcome.

Own document parsing: turning supplier-uploaded invoices, catalogs, and contracts into structured context the agent can reason over.

Wrap Coupa supplier APIs as agent-callable CLI tools, with clean error surfaces and sensible defaults for an agent caller.

Ship supplier-facing skills on top of the harness: the procedural instructions and tool compositions that let the agent handle specific tasks end-to-end.

Debug messy production behavior: why did the agent take that path, where did it get confused, and what tool, context, or harness change fixes it?

Partner with the senior engineer on harness architecture calls as you find gaps while working across the stack.

Have 3–5 years shipping production software and are strong in Python (TypeScript a plus).

Have shipped at least one LLM-powered feature or product, even if small, and can talk about what went wrong and what you'd do differently.

Are a daily, heavy user of agentic coding tools — Claude Code, Cursor, Codex, or equivalents.

Have side projects. Real ones. Things you've built because you couldn't stop thinking about them. Ideally things the new generation of AI tools made possible for you to finish.

Write and speak English fluently. Skill authoring is prose-heavy and the team operates in English.

Debug by reading, not by guessing. You reach for logs, traces, and evals before theorizing.