Long Thành Software

SaaS platform · AI

QKIT NEXUS

A multi-tenant SaaS platform for AI tooling, content management and business automation.

Projects

Overview

QKIT NEXUS is a multi-tenant SaaS platform, built so that a company can run AI tooling and operational management on one system rather than buying separate products and gluing them together.

The problem is that the feature areas are very different from one another — AI chatbot, knowledge base, content management, marketing automation, customer management — while all of them need the same foundation: which tenant is this, what is this user allowed to see, and how is one tenant's data kept from ever reaching another. Most of the design work sits in that foundation rather than in any individual feature.

The system runs on a multi-tenant architecture with multi-level authorization, scales per tenant and per workspace, and ships an interface in three languages: Vietnamese, English and Hebrew.

Projects

Scope of work

  • Tenant and workspace management
  • User, role and permission management
  • AI chatbot and knowledge base on a RAG architecture
  • Content management and blog publishing
  • Marketing workflows and lead management
  • Real-time chat
  • External AI provider integration over API and webhooks
  • Chatbot widget embeddable in third-party sites

Role

Responsibilities

Project Leader / Lead Full-stack Engineer

  • Business requirements analysis, BRD and PRD
  • System architecture and multi-tenant architecture design
  • Backend and frontend dashboard implementation
  • Authentication, authorization, role and permission design
  • Tenant and workspace data isolation
  • Mentoring backend developers and reviewing implementation
  • Production deployment, performance and scalability work

Projects

Key features

  • Tenants and workspaces

    Each tenant holds its own data space, subdivided into workspaces. Authorization is checked at both levels, so a user who belongs to several workspaces still sees only what they were given.

  • AI chatbot on a private knowledge base

    Each tenant's documents are embedded and retrieved through a RAG architecture, so answers come from that tenant's own material rather than from the model's general knowledge.

  • Authentication and authorization

    JWT with refresh tokens, configurable roles and permissions, and every data query passing through a tenant check before it reaches the database.

  • Real time and background work

    Chat and notifications run over Socket.IO; long-running work — document embedding, bulk sends — is queued rather than held open on a user request.

  • Embeddable widget

    The chatbot is packaged as a widget for third-party websites. It runs in an environment nobody controls, so it has to scope itself and handle connection failure on its own.

  • Content and marketing

    Authoring and publishing, lead management and marketing workflows sit on the same authorization layer as the rest of the platform.

Architecture & technologies

Engineering challenges

  • Tenant data isolation

    The largest risk in any SaaS platform: one query missing its tenant condition is a data-leak incident. Isolation is pushed down into the data-access layer and a split Prisma schema, rather than relying on every developer remembering a condition at every call site.

  • Two-level authorization

    Tenant-level and workspace-level permissions can contradict each other. The permission model has to answer clearly which one wins — and answer identically in the backend and in the interface.

  • RAG retrieval quality

    A chatbot is only useful when it retrieves the right passage. The hard part is not calling a model; it is chunking, embedding and ranking across tenants whose document quality varies enormously.

  • Scaling per tenant

    Load is not evenly distributed: one large tenant can outweigh a hundred small ones. The backend is split into modules communicating over events so the loaded part can scale on its own.

Projects

Outcome

  • A platform running on multi-tenant architecture with data isolation and two-level authorization
  • An AI chatbot answering from each tenant's own knowledge base through a RAG architecture
  • An interface in three languages: Vietnamese, English and Hebrew
  • The chatbot packaged as a widget embeddable in third-party sites

Projects

Project images

  • Tenant dashboard
  • AI chatbot and knowledge base
  • Workspace and permission management

Services

Related services

← All projects

Start here

Ready to launch?

Send a few lines about what you need, or just call. We answer within one business day and the first consultation is not billed.