AI Agent Infrastructure Engineer (or Team) | Repeatable Private VPS Builds for B2B Clients

We build private AI agent infrastructure for small and mid-size businesses in USA. Each client gets a fully isolated VPS deployed as a repeatable standard build. We have a growing pipeline and need an engineer or small team who can own the technical builds on an ongoing basis at fixed pricing. What the infrastructure needs to do: Server management, relational database, object storage, a central API gateway routing to Anthropic Claude and Google Gemini with per-agent routing and spend tracking (clients own their own API keys), an automation engine for scheduled workflows and integrations, standalone tool apps each running at their own URL, uptime monitoring, and HTTPS with full client isolation and revocable access. What sits on top: AI agents are room-based conversational assistants with per-agent memory/rules, client-curated knowledge bases, and automation integrations. Rooms segment by function. Users can share rooms or have independent memory under the same agent. Agent types include CRM copilots, lead research, email triage, meeting prep, content drafting, project tracking, deal scoring, and knowledge base retrieval. Custom tools are standalone web interfaces: document processors, proposal builders, dashboards, intake forms, route dispatchers, and more. Each connects to the same database and storage as the agents. In your proposal, address all of the following: Recommended Stack and Chat Interface. Tell us what you would build this on and why. We have concerns about Open WebUI. The chat interface must support multi-user auth, role-based access, per-room knowledge bases, memory/rules configuration, and a clean UI for non-technical users. Recommended VPS Hosting. We default to Hetzner. Better option? Include reasoning. Fixed Pricing. Provide a fixed price for each: • Base infrastructure setup (full environment, ready for agents and tools) • Build one new AI agent (room, integrations, workflows, knowledge base, memory config) • Build one new custom standalone tool • Replicate an existing agent for a new user (separate memory, same structure) • Add a new room within an existing agent (shared memory, new context) • Add a new user to an existing room (provisioning only) • Hourly rate for post-build support and modifications Team and Capacity. Team size and roles. Concurrent build capacity. Turnaround for full setup vs. single agent/tool. Availability and earliest start. Relevant Experience. Self-hosted AI infrastructure, agents with RAG and persistent memory, Claude/Gemini API, standalone tool deployments, multi-client isolated environments. Security. Clients include law firms and regulated businesses. Describe your approach to data isolation, access control, and credential management. Engagement model: Fixed price per build type, hourly for post-build work. We handle client communication, scoping, and project management. You get a defined spec. Expect multiple builds per month.