AI Platform Engineer, Kraken

• Enterprise SaaS administration at 1,000+ seats, ideally including an AI or LLM platform
• Demonstrated ability to build and ship working software, not just configure vendor products. Python or a comparable language should be natural to you. Version control, CI, containers, and secrets management are everyday tools
• Experience integrating against LLM and SaaS APIs with working familiarity of agent and tool use patterns, including MCP or equivalent tool calling architectures
• Working knowledge of identity and access. Okta or an equivalent IdP, OAuth, SCIM, SSO, and federated machine identity are things you've worked with
• Operational discipline. You understand intake queues, SLAs, runbooks, escalation hygiene, and the instinct to document a fix the first time you make it
• Genuine teaching ability. You can explain a technical capability three different ways on the same day to compliance, traders, and backend engineers
• Clear written communication for technical documentation, runbooks, and policy pages
• Judgment about what to build. You can tell the difference between a capability the organization actually needs and a demo that will be abandoned in a month

Desirable

• Design and build the technical layer of the AI platform, from integration surface to production runbook. Identify capability gaps between what's possible and what the organization can safely use, then close them by evaluating vendors or building internally
• Own the governed integration gateway as an engineering surface. You'll handle connector onboarding, authentication flows including OAuth and federated identity, reliability, and vendor escalation when things break
• Build internal tooling that extends the platform. Think agent scaffolding, reusable skills and prompt assets, evaluation harnesses for model changes, provisioning automation, and integrations between AI tooling and your systems of record
• Configure and operate tiered spend controls across the AI estate. Set caps by role and tier, build alerting rules, manage exception queues, run monthly reconciliation against vendor commitments, and eliminate the current single author risk in cap automation
• Implement the technical controls behind policy on data handling, retention, and model eligibility. This means identity integration, SSO and SCIM, entitlement by role and tier, and workspace configuration where retention terms differ by model class
• Work directly with teams to unblock them. Sit with them, understand the workflow, build or configure the thing that enables them to get real leverage from the tools
• Run technical enablement. Deep dives for tool rollouts, office hours for model releases, and authority over the engineering detail in knowledge base material
• Administer the enterprise AI estate end to end. Model turn ups, feature enablement, deprecation, provisioning queue, vendor configurations, and seat true ups

• Unless a specific application deadline is stated in the job posting, applications are accepted on an ongoing basis