AI Engineer, Agentic Core

Medra; Stati Uniti; Remoto; USD 120.000 - 180.000; Tempo pieno

Descrizione del lavoro

Medra is building the shared intelligence layer behind conversational and task-oriented AI agents for customer and internal use. The role covers multi-agent architecture, evaluation, reliability, and model-provider choices across a large multilingual workload spanning 19 countries.

You will develop orchestration, improve reliability, and create robust evaluation frameworks while guiding reusable patterns into core systems. Fluency in Spanish for prompts and evaluations is required.

Competenze

• Shipping production software experience (3+ years), including recent work with LLMs.
• Strong Python or TypeScript skills.
• Experience building agentic systems with tool use, multi-step workflows, or autonomous decisions.
• Experience operating non-deterministic software in production and diagnosing its behavior.
• Fluency in Spanish for writing prompts and evaluations.
• Comfort working through ambiguous problems without detailed specifications.

Mansioni

• Develop multi-agent orchestration, including supervisor patterns, specialist agents, routing, and state management.
• Improve reliability across hundreds of thousands of conversations.
• Create and strengthen evaluations that show whether agent behavior is improving.
• Define how shared frameworks connect with client-specific work, promoting reusable patterns into core systems.
• Help guide model and provider choices across a system that uses multiple providers.
• Debug non-deterministic behavior using traces, guardrails, and fallback strategies.

Conoscenze

Python

TypeScript

LLMs experience

Agentic systems

Non-deterministic software

Spanish fluency

Ambiguity tolerance

Strumenti

LangSmith

WhatsApp Business API

Descrizione del lavoro

Role overview

Build the shared intelligence layer behind conversational and task-oriented AI agents for customer and internal use. The role covers multi-agent architecture, evaluation, reliability, and model-provider choices, with the aim of turning useful solutions from client work into reusable capabilities. These agents operate across a large, multilingual conversation workload spanning 19 countries.

Responsibilities

• Develop multi-agent orchestration, including supervisor patterns, specialist agents, routing, and state management.
• Improve reliability across hundreds of thousands of conversations.
• Create and strengthen evaluations that show whether agent behavior is improving.
• Define how shared frameworks connect with client-specific work, promoting reusable patterns into core systems.
• Help guide model and provider choices across a system that uses multiple providers.
• Debug non-deterministic behavior using traces, guardrails, and fallback strategies.

Requirements

• At least three years shipping production software, including substantial recent work with LLMs.
• Strong Python or TypeScript skills.
• Experience building agentic systems with tool use, multi-step workflows, or autonomous decisions.
• Experience operating non-deterministic software in production and diagnosing its behavior.
• Fluency in Spanish for writing prompts and evaluations.
• Comfort working through ambiguous problems without detailed specifications.

Nice to have

• Experience with evaluation tools such as LangSmith or similar frameworks.
• Voice agents, telephony, latency optimization, or messaging platforms such as WhatsApp Business API.
• Multi-tenant design experience, especially systems configured by other engineers rather than copied and modified.