Platform Engineer: AI Agent Harness & Orchestration

Kyndryl

Madrid

Híbrido

EUR 70.000 - 100.000

Jornada completa

14 días+

Recibe más respuestas de empleadores

Envía un currículum específico para el puesto de trabajo en cuestión de minutos.

Ventajas ofrecidas por este puesto de trabajo

Be Well program

Hybrid-friendly culture

Learning opportunities (certifications

Career development

Descripción de la vacante

Kyndryl in Madrid is seeking a Platform Engineer to build the internal agent harness: the runtime, orchestration, and guardrails that turn foundation models into reliable AI agents. You will treat the harness as a product used by engineering teams, delivering measurable agentic productivity and handling model upgrades without rewrites.

The role demands hands-on Python, Kubernetes, IaC, and cloud experience, plus professional English and Spanish.

Formación

5+ years of Platform Engineering, DevOps or SRE experience.
Hands-on experience building agent harnesses or agentic systems.
Proficient Python software engineering for production systems.
Strong cloud-native background: Kubernetes, containers, IaC, and one major cloud (AWS, Azure or GCP).
Understanding of how LLM agents fail and how to make them reliable.
Experience treating a platform as a product with contracts, SLOs, adoption metrics and real users.
Professional working English and Spanish.

Responsabilidades

Design and build the agent harness: runtime, orchestration, tool and MCP integration, context and memory, evaluation, guardrails and observability.
Build golden paths and reusable patterns so teams ship agents that are reliable, governed and measurable.
Define the oversight model: policy‑as‑code, human‑in‑the‑loop checkpoints, loop and failure detection, evaluation harnesses, and cost control.
Integrate agents with the systems they act on: CI/CD, infrastructure, cloud services, observability and ticketing, through APIs and MCP.
Make agentic productivity measurable: define what productive means for each use case, instrument it, and run the harness on that data.
Keep the harness model‑agnostic and versioned, so it survives model upgrades and provider changes without rewrites.
Mentor teams, run enablement, and reduce friction between Dev, Ops, Security and the people adopting agents.