Platform Hardening & Ai Agent building

XOps is a B2B commercial intelligence platform that detects buying signals across target accounts, synthesizes an account thesis (why this account, why now, what angle), and maps the warm path to the decision-maker. The platform is in production with paying customers. We are bringing on a senior engineer to harden the two core intelligence systems and to complete the signal-resolution agents that determine whether the right things surface for the right accounts. This is a correctness-and-reliability engagement, not a greenfield build. The architecture exists. The patterns exist. Your job is to make the existing systems production-hard and to build new agents that conform to those patterns, not to introduce a new stack. The stack you'll work in Node.js on Fly.io Neon PostgreSQL (large existing schema, 60+ tables) A multi-agent pipeline with an observe / diagnose / approve operational boundary A connector-based architecture (SignalOS) and a graph-based access/relationship layer (Access Map)