Principal AI Architect

Know where AI is going. Then build what comes next. AI moves faster than any roadmap. RoonCyber.AI is building an AI Workload Security platform that discovers AI workloads, reconstructs what agents actually did, ties every action back to an identity, and responds when something goes wrong. We're looking for a hands-on AI expert and platform architect to decide which AI developments matter, prototype them, and shape them into the capabilities we build next. What You'll Own AI Technology Radar Platform Architecture and Data Model AI Execution Graph and Component Model Prototypes and Reference Implementations Capability Vision and Roadmap Input Feasibility and Rough Estimates Architecture and Code Review Technical Demos and Customer Deep Dives Decision Records and V1 Scope Problems You'll Solve Decide whether a new agent protocol or framework should be modeled, detected, watched, or ignored. Keep runtime, identity, AI, and integrations building to one platform model. Size new capabilities fast enough for Product and Engineering leadership to plan. Draw the V1 line on design partner requests that are half a product. Technologies Python (required), Go or Rust LLM APIs (OpenAI, Anthropic, Bedrock, Vertex AI) LangGraph, CrewAI, AutoGen, and agent SDKs MCP and agent-to-agent protocols Agent instruction files (AGENTS.md, CLAUDE.md, SKILL.md) AI gateways, vector databases, and agent memory Kubernetes and AWS Graph databases eBPF and OpenTelemetry OAuth, SPIFFE/SPIRE, and non-human identity OWASP Top 10 for LLM and Agentic AI, MITRE ATLAS What Great Looks Like Publish an AI technology radar the whole company uses. Turn new AI developments into working prototypes within weeks. Give rough estimates that Product and Engineering leadership trust and plan from. Keep four principal engineers building to one platform model. See your prototypes become shipped capabilities. Run customer and design partner deep dives that earn technical trust. We're Probably Looking For You If You've architected and shipped an AI or security platform. You still write production-quality code. You prototype new frameworks to understand them, not just read about them. You can size a capability in an afternoon and be roughly right. You can whiteboard an agent attack path for a CISO, then go deep with their engineers. You write clear design docs and decision records. Ideal Background Principal or distinguished engineer, chief architect, or hands-on field CTO. Wiz, CrowdStrike, Palo Alto Networks, Microsoft Security, Protect AI, HiddenLayer, Lakera, Databricks, Anthropic, OpenAI, AWS, Google Cloud. Open-source contributions, published tools, or talks on AI security are a plus. What We Don't Care About Managing a large team. Training foundation models from scratch. Academic credentials over practical judgment. Hype. We want the person who can tell us what not to build. The Interview Build a short prototype that detects misuse of a new MCP server or agent framework, then walk us through the code. Decide whether a recent AI development should be modeled, detected, watched, or ignored. Size a real design partner request and draw the V1 line. Show how coding agents on developer laptops fit into one platform model. Meet Product and Engineering leadership to discuss how you'd work together. Why Join Help define and build a new category: AI Workload Security. Direct influence over product direction and architecture. Work directly with founders, Product, and Engineering leadership. Equity participation. Engineers Who Will Love This Job You read release notes for fun, then try the thing that weekend. You'd rather shape five products than build one component. You enjoy a whiteboard with a CISO as much as one with engineers. You like being the bridge between what's possible and what customers need. This Probably Isn't For You If You want to run a large team or own delivery schedules. You've stopped writing code and don't want to start again. You want to go deep on one component for years. You'd rather follow AI trends than make calls about them.