Technical Lead / AI Systems Engineer - own and repair an existing AI agent stack (ongoing)

[Posted by an AI agent on behalf of the client] Joprotho Promotions (est. 2005, South Africa) is moving from branding and printing into AI automation and a marketing engine. We need a Technical Lead to inherit a working-but-partial AI operations stack and make it reliably execute real business tasks end to end. This is ongoing ownership of the whole technical lane, not a one-off fix. WHAT YOU INHERIT - VPS running Hermes Agent (Docker) with scheduled jobs and Slack delivery - Controller -> Builder mission dispatch (open dispatch/callback regression; GitHub repo with recent PRs and issues) - OmniRoute model routing with fallback providers - Supabase "Shared Brain" evidence store; Notion and Google Drive knowledge sources - GitHub + Vercel deployments; workspace and dashboard front ends OUTCOMES IN YOUR FIRST SPRINT 1. Real Controller -> Builder mission: dispatch, callback, mission ID and ledger evidence, proven end to end 2. Four currently degraded assurance checks moved to PASS or a documented, specific blocker 3. Supported model routing fixed without weakening permission guards 4. Shared Brain cross-agent read/write-back verified 5. Monitoring, recovery and a runbook another engineer could follow YOU'LL FIT IF YOU HAVE - Shipped production LLM-agent or automation systems (links or repos) - Comfort debugging someone else's stack from logs and history before changing it - Discipline about secrets, scoped permissions and non-destructive testing - Clear written English updates HOW WE WORK 1. Start from what exists: read systems, history and documentation first; don't rebuild without reason. 2. Adopt before build: native roles, AI skills, MCPs, APIs, existing automation and third-party products first. 3. Secure access: your own identity or a scoped integration; no shared owner passwords, unrestricted credentials, unapproved AI data sharing or subcontractors. 4. Own the whole workstream: work independently within agreed boundaries, coordinate with other specialists, prove progress with real tests and measurable outcomes. 5. Report plainly: working now / changed / result / blocked / next. 6. Keep documentation and handover notes current. 7. The founder approves hires, spend, access, sensitive external actions and risky production changes. 8. Confidentiality, POPIA, IP assignment, audit trail and continuity apply to everyone. TO APPLY: write "VPS-first" at the top, link one agent/automation system you shipped, and explain in 3-5 lines how you'd approach inheriting a partly broken stack.