Lead AI Engineer: Agent Architect for Production AI

Coupa; Ciudad de México; Presencial; MXN 600.000 - 950.000; Jornada completa

Descripción de la vacante

Coupa in Mexico City is seeking a software engineer to ship production software and own critical pieces of the agent harness. You will build and extend the eval pipeline, manage context for long-horizon tasks, design sub-agent patterns, and convert supplier data into actionable context for the agent.

You will work with Python, LL-powered features, and agentic tools while collaborating in English. This role emphasizes debugging, logs analysis, and iterative improvement in a fast-paced environment.

Formación

• 3–5 years shipping production software with Python; TypeScript a plus.
• Experience shipping at least one LLMed powered feature or product.
• Daily heavy use of agentic coding tools such as Claude Code, Cursor, Codex.
• Real side projects driven by AI tools.
• Fluent English writing and speaking; prose-heavy skill authoring in English.
• Debug by reading logs, traces and evals before theorizing.

Responsabilidades

• Build and extend the eval pipeline: measure task completion, catch regressions, and fix production failures.
• Work on context management: determine what the agent sees, when, and how long-horizon tasks stay coherent.
• Design sub-agent patterns: decide when to fan out and how to compose specialized agents.
• Own document parsing: convert supplier-uploaded invoices, catalogs, and contracts into structured context.
• Wrap Coupa supplier APIs as agent-callable CLI tools with clear error surfaces.
• Ship supplier-facing skills on top of the harness for end-to-end tasks.
• Debug messy production behavior by analyzing paths, then applying fixes.
• Partner with the senior engineer on harness architecture calls.

Conocimientos

Python

TypeScript

LLM-powered features

Agentic coding tools

English fluency

Debug logs/evals

Herramientas

Claude Code

Cursor

Codex

Descripción del empleo

Coupa in Mexico City is seeking a software engineer to ship production software and own critical pieces of the agent harness. You will build and extend the eval pipeline, manage context for long-horizon tasks, design sub-agent patterns, and convert supplier data into actionable context for the agent.

You will work with Python, LL-powered features, and agentic tools while collaborating in English. This role emphasizes debugging, logs analysis, and iterative improvement in a fast-paced environment.