Senior Systems Engineer, Enterprise AI Platforms

Senior AI Platform EngineerCoreWeave is The Essential Cloud for AI™. Built for pioneers by pioneers, CoreWeave delivers a platform of technology, tools, and teams that enables innovators to build and scale AI with confidence. Trusted by leading AI labs, startups, and global enterprises, CoreWeave combines superior infrastructure performance with deep technical expertise to accelerate breakthroughs and turn compute into capability. Founded in 2017, CoreWeave became a publicly traded company (Nasdaq: CRWV) in March 2025.CoreWeave is hiring a Senior AI Platform Engineer to own the design, implementation, and governance of AI across CoreWeave's employee environment. You'll be the technical owner for the AI tools our employees use every day, on every surface they ship on: chat assistants, desktop apps, coding agents, AI terminals, AI browsers, workplace knowledge tools, and whatever comes next. You'll make them safe to adopt by building CoreWeave's guardrails directly into the platform through identity, device, and tenant-level controls rather than policy documents alone.This role is for someone who is genuinely passionate about AI and lives on its bleeding edge: someone who is already tuning agent behavior every week, has run swarms of agents and agent loops rather than just chatted with a model, and wants to keep learning at the velocity this field now moves. The tools, models, and best practices in this job will change monthly. That should excite you, not exhaust you.This is not a traditional AI/ML engineering role, and it is not a traditional IT administration role. It is deliberately a composite role that combines three things: production software and SRE engineering, hands-on LLM and agent work, and enterprise IT controls across identity, devices, and SaaS. We expect strong candidates to arrive deep in two of the three and grow into the third. Tell us which two.What success looks like in the first year: CoreWeave employees get frontier AI capabilities before almost any enterprise on earth. When a vendor ships a new model, agent surface, or feature, you already know it's coming; you're plugged into vendor roadmaps, betas, and the community and you have an evaluated, governed rollout ready in days, not quarters. That speed is possible because of the foundation you build in year one: sanctioned tools governed through identity, a connector and MCP allowlist in production, an intake process so "is this allowed" has an answer, internal agents shipping through an IT-owned gateway, and leadership visibility into usage, cost, and risk. The foundation is the floor; the pace is the job.AI platform lifecycle and governanceOwn the end-to-end lifecycle of employee-facing AI platforms across every surface they ship on (web, desktop, mobile, IDE, terminal, and browser): evaluation, procurement support, tenant design, rollout, configuration management, and ongoing operations.Keep the Platforms in Scope list current as the market moves.Establish CoreWeave's AI operating framework for IT: intake and review for new AI tools and agents, standards for internal bots, runbooks, a clear path from pilot to supported service, and enablement content that helps employees use AI tools well and safely.Guardrails as enforceable controlsDesign and implement AI guardrails as enforceable controls, in partnership with Enterprise Security and IT Identity and Access: SSO and SCIM provisioning, role- and group-based feature entitlement, data-retention and training opt-out settings, browser-extension and desktop-app policy, and device-trust requirements for access.Build and operate connector and MCP allowlists with data-source scoping across Okta, Google Workspace, Slack, Atlassian, GitHub, ITSM, and HRIS.Partner with Enterprise Security, Legal, and Privacy to translate acceptable-use, data-classification, and regulatory requirements into technical policy, and support audits with the evidence those controls produce.Infrastructure you'll ownThe IT-owned control layer for CoreWeave's internal chatbots and agents: the gateway, identity, secrets, MCP allowlisting, logging, rate limiting, and cost attribution in front of the serving stack, deployed on CoreWeave infrastructure alongside the teams that run it.Inference endpoints, agent hosts, and developer AI tooling on managed devices, treated as managed endpoints with a lifecycle, configuration baselines, and observability.The IT-owned share of CoreWeave's internal MCP servers, reusable agent skills, and shared agent configuration (system prompts, AGENTS.md and CLAUDE.md instruction files, tool policies), plus the identity, logging, and allowlisting standards, with reference implementations, that agents and MCP servers built by other teams must meet.Operations, reliability, and telemetryHold the AI services you own to SRE standards: automation and infrastructure-as-code for everything repeatable, reliability targets, on-call, and incident response.Instrument usage, cost, quality, and risk telemetry across AI platforms, and build the dashboards and alerting that show leadership what is being used, by whom, and at what spend.Serve as the senior escalation point for AI tooling issues.Communicate clearly and proactively with executives, cross-functional partners, and employees.Who You Are:6+ years in software engineering, site reliability engineering, or IT systems engineering where you built and ran things in code, not just configured them.2+ years of hands-on experience building with or administering LLM-based products: API integration, prompt and tool design, retrieval pipelines, agent frameworks, multi-agent workflows and agent loops, or enterprise administration of tools such as Claude, ChatGPT Enterprise, Copilot, Gemini, Glean, or Rovo. You have shipped or run something with these, not just used them.Strong software engineering skills in Python, Go, Rust, or TypeScript: you have built services and integrations that other people depended on and kept them running.Experience operating services in production (containers, infrastructure-as-code, CI/CD, secrets management, and observability), or you have automated enterprise systems with code (Terraform for Okta or Google Workspace, CI/CD for device and SaaS configuration, API integrations a company depended on) and are ready to go deeper on containers and observability.A working understanding of how enterprise IT control systems function, even if you have not administered them yourself: Identity and access: how SSO, SCIM, SAML/OIDC, and OAuth grant, scope, and revoke access, and what Conditional Access, Device Trust, and Zero Trust mean in practice.Endpoint and device management: how MDM and device-trust tooling enforce policy on a managed laptop or browser.SaaS administration: how permission and data-sharing models in tools like Google Workspace, Slack, Atlassian, and GitHub decide who can see what.Proven ability to produce clear technical documentation and lead complex, cross-functional technical initiatives.Preferred:Real-world experience administering one or more of these at enterprise scale: identity (Okta or Entra ID), endpoint management (Apple MDM/DDM, Jamf, Kandji, Munki, Intune, Chrome Enterprise), or SaaS platforms (Google Workspace, Slack, Atlassian, GitHub).Experience deploying and governing enterprise AI tools across a 1,000+ person organization.Hands-on experience building custom Model Context Protocol (MCP) servers and integrations, authoring reusable agent skills, and shaping agent behavior through system prompts and repository-level instruction files (AGENTS.md, CLAUDE.md, or equivalent).Working knowledge of data governance for AI: data classification, retention, training opt-out, DLP, and the audit and logging capabilities of the major AI vendors.Experience self-hosting and serving models.Familiarity with compliance frameworks such as SOC 2, ISO 27001, SOX, and GDPR, and with emerging AI governance standards such as NIST AI RMF and ISO/IEC 42001.Wondering if you're a good fit?You may be a great fit if you already run swarms of agents and agent loops for your own work; you've been the person who figured out how to say "yes, safely" to a new AI tool instead of "no"; you're curious about how an Okta policy or an MDM profile