Multi-Tenant Real Estate Operating System

Harlem Manage

Built the full backend of a multi-tenant proptech platform: 46-tool role-based AI agents, real-time M-Pesa WebSocket payment flows, a 3-queue Celery task system, and a PII-scrubbing observability stack.

AI agent tools

RBAC user roles

Celery queue types

Problem statement

Property management in the target market is not just CRUD over properties and tenants. It requires organization-level isolation, financial correctness, payment reconciliation, operational coordination across multiple roles, and localized workflows that reflect how real rental systems run in Kenya.

Architecture breakdown

I framed the system as a distributed, multi-tenant SaaS platform with strict relational modeling, RBAC enforcement across the application and AI layer, M-PESA-first payment design, and Akoko as a first-class intelligence engine with role-specific execution modules, channel-aware outputs, and human escalation paths.

- Role-based AI agent system (GPT-4.1 Responses API) with dynamic tool selection across 4 roles and 46 tools â€” full audit trail per message for compliance
- M-Pesa C2B STK Push with OAuth2, field-level encrypted credentials (per organisation + property), and live payment status streamed via JWT-authenticated WebSocket consumers over Redis
- 3-queue Celery architecture over RabbitMQ (notification / scheduled / background) with 5 Beat tasks: monthly ledger generation, maintenance automation, payment reminders
- Production observability: structured JSON logging, per-request UUID tracing, recursive PII scrubber redacting 20+ sensitive field patterns before Sentry/GlitchTip transmission
- 2-level multi-tenancy (Organisation â†’ Property) with fully isolated payment flows, RBAC enforcement, and field-level encryption transparent at ORM level

Tech stack explanation

Next.jsDjangoDjango REST FrameworkPostgreSQLM-PESAWhatsApp/SMSAkoko AI

System diagram

[ Organizations ]
      |
      v
[ Multi-Tenant Core ] ---> [ Properties / Units / Tenants ]
      |                             |
      |                             +--> [ Lease & Rent Engine ]
      |                             +--> [ Ledgers / Reconciliation ]
      |
      +--> [ RBAC Layer ]
      |        +--> Owner
      |        +--> Manager
      |        +--> Caretaker
      |        +--> Tenant
      |        +--> Guard
      |
      +--> [ Akoko Intelligence Layer ]
               +--> Owner Execution
               +--> Caretaker Execution
               +--> Tenant Execution
               +--> Guard Execution

Key challenges

Harlem Manage is a Kenya-first, multi-tenant real estate operating system built for landlords, agencies, and property firms. It combines property workflows, tenant and lease management, financial reconciliation, communication channels, and Akoko, a deployed operational intelligence layer that adapts by role.

- Positioned the product as a real estate operating system for African rental infrastructure, not a simple management panel.
- Integrated operational intelligence directly into the product through Akoko rather than treating AI as a future add-on.
- Created a strong systems case study spanning SaaS architecture, fintech workflows, and access-controlled AI behavior.

What I learned

AI becomes much more valuable when it is embedded inside operational workflows rather than exposed as a generic assistant.

Multi-tenant architecture needs to be treated as a system-wide discipline across schema design, service logic, and permissions.

Local market context like M-PESA, language behavior, and escalation norms can meaningfully shape architecture decisions.