AI Platform Lead

The ICRC is seeking an AI Platform Lead to own the hybrid AI platform that underpins the organization's centralized, transversal AI services (e.g. translation, entity recognition, specialized and generic LLMs). The platform covers AI services deployed on Kubernetes, model deployment on GPUs, the AI gateway and the self-service capabilities offered to advanced users. The AI Platform Lead defines the platform vision and roadmap, leads its engineering and operation and manages its backlog. This enables product teams, data scientists and engineers to consume and build AI solutions securely and reliably in support of the ICRC’s mandate of protecting and assisting victims of armed conflict and other situations of violence. The role requires a good understanding of cloud-native infrastructure, GPU and LLM serving and open-source AI platform technologies. Main Duties & Responsibilities Own the AI platform vision, roadmap and backlog. Prioritize features, improvements and technical debt with product owners, AI teams and other stakeholders, balancing user needs, security, cost and capacity. Manage the deployment and operation of centralized, transversal AI services (e.g. translation, entity recognition, specialized and generic LLMs) on the hybrid Kubernetes platform, making them reusable across ICRC products. Work closely with the Kubernetes product team on workload

, namespaces, resource quotas and release practices. Manage model deployment on GPU infrastructure. Serve open-source and third-party models (LLMs and other ML models) with appropriate inference engines, covering GPU scheduling and sharing, capacity planning, autoscaling, performance tuning and model release and rollback. Co-design and operate the AI gateway (e.g. LiteLLM or equivalent) with the security and network teams responsible for perimeter security and access controls. Run it as a single governed access point for model delivery (routing, fallback, access to internal and external models), Model Context Protocol (MCP) servers and tools, consumption monitoring (usage, quotas, cost tracking, logging) and guardrails. Design and operate self-service capabilities (e.g. OpenShift AI or comparable solutions) that enable advanced users to develop, train and deploy models in a governed environment, including workbenches, pipelines, model registry and resource quotas. Provide shared AI building blocks, such as vector and graph databases, agent components, MCP servers, model registry and experiment tracking, as documented services.