Founding architect

A week at Navabi.

• Design and record the target architecture: boundaries between services, the data model and the choices that are expensive to reverse later. You write them down as decisions, with the reason attached.
• Multi-tenancy as the foundation: every customer separated down to the keys, with isolation per cell so that growth and outages never reach other customers, and data that stays in the customer's region.
• The chain from AI draft to lawyer to signed off as a durable workflow: long-running matters with human steps, retries and idempotency, and an audit log with a hash chain that cannot be altered afterwards.
• A model gateway that is not tied to one vendor: routing per task, redaction of sensitive data, cost per credit per customer, and evaluations that block any release if quality drops.
• One API for everything: the portal, Slack, Teams, the MCP server and customers' agents use the same versioned API, with keys, limits and webhooks.
• Security and certification from the first line of code: infrastructure as code, SSO, role-based access, audit logs and the evidence that ISO 27001, SOC 2 and the EU AI Act require.

• At least six years of experience as an engineer, part of it on the architecture of a multi-tenant SaaS product that grew substantially.
• You have put an LLM application into production, with real users and evaluations.
• Deep knowledge of Postgres, event-driven design and durable workflows.
• You know what to build now and what to defer deliberately, and you can explain that choice to an investor or an auditor.
• Experience with security and compliance in a regulated sector is a strong plus.