Full-Stack SaaS · AI Automation · Production

SHAI For AI — AI Training Platform

End-to-end SaaS platform built from scratch for an AI training and certification company in Amman, Jordan. Custom CRM, bilingual AI sales bot, Discord enrollment automation, and a full React/Node.js production stack. Sole engineer responsible for architecture, development, deployment, and operations.

Sole Engineer Architecture Backend API Frontend (React) Database Design AI Bot Integration Discord Automation CI/CD Deployment Production Ops

Problem

SHAI For Artificial Intelligence needed a complete digital platform to support course sales, student enrollment, lead management, and post-sale support — with no existing system in place. Operational requirements included multilingual support (Arabic and English), automated lead qualification, structured course enrollment workflows, and an internal CRM for the sales team.

Off-the-shelf CRM and LMS platforms were evaluated and rejected: none satisfied the bilingual AI requirements, the custom enrollment flow, the Discord-based course delivery model, or the budget and control constraints of a startup.

Solution

Built a full-stack SaaS platform from scratch. The system has three primary layers: a customer-facing marketing and registration site, an internal CRM and admin dashboard for the sales and operations team, and an AI-powered sales bot that handles inbound leads in Arabic and English before routing them into the CRM pipeline.

A Discord automation layer handles post-enrollment course delivery: once a student enrollment is confirmed in the CRM, the system automatically creates a private Discord server channel, assigns the student role, and sends onboarding materials — eliminating manual steps from the operations team's workflow.

Tech Stack

Frontend

React 18ViteTailwind CSSJSX

Backend

Node.jsExpress.jsJWTHelmetRate Limiting

Database

Prisma ORMSQLite

AI / Automation

LLM APIIntent ClassificationDiscord.js v14node-cronnodemailer

Infrastructure

DockerRailway (nixpacks)VercelCI/CD

Architecture

React FrontendReact 18 with Vite and Tailwind CSS. Public-facing marketing and registration pages. JWT-secured admin dashboard for the CRM and operations team. Deployed on Vercel with CI/CD auto-deploy from main branch.

Node.js APIExpress.js REST API. JWT authentication with role-based access control (student, admin). Helmet security hardening, request rate limiting, and input validation. Containerised with Docker, deployed on Railway via nixpacks.

ShaiCRMCustom-built CRM using Prisma ORM with a SQLite data layer. Models: Registration, User, Course, Enrollment, SupportTicket. Full lead lifecycle from initial intake through enrollment confirmation, with admin dashboard for the sales team.

AI Sales BotBilingual Arabic/English AI bot handling inbound lead qualification. Uses LLM API for intent classification: enquiry routing, course recommendation, FAQ handling, and lead capture. Pushes qualified leads directly into the CRM pipeline via internal API calls.

Discord AutomationDiscord.js v14 bot triggered on CRM enrollment confirmation events. Auto-creates a private course channel, assigns student role, delivers onboarding materials, and sends a structured welcome message. Eliminates all manual Discord operations for the team.

Scheduled Jobsnode-cron handles recurring tasks: enrollment reminders, follow-up sequences, session notifications. nodemailer manages transactional email delivery for registration confirmations and support ticket updates.

Key Features

Bilingual AI bot — Handles inbound leads in Arabic and English with intent classification and CRM routing
Custom CRM — Built from scratch with Prisma; full lead lifecycle, enrollment tracking, and admin dashboard
Discord enrollment automation — Zero-touch channel creation and role assignment on enrollment confirmation
JWT auth + RBAC — Secure admin access with role-based dashboard views
Security hardening — Helmet, rate limiting, input validation, and environment secret management
CI/CD pipeline — Auto-deploy on push to main; Railway (backend) + Vercel (frontend)
Scheduled automations — Enrollment reminders, session alerts, follow-up sequences via node-cron

Deployment

Full production deployment from day one:

Railway — Backend API (nixpacks) Vercel — React Frontend (CI/CD) Docker — Containerised backend Discord API — Bot hosting Environment secrets — Railway + Vercel vaults

Security & Privacy

JWT-based authentication with role-based access control separates student, admin, and system-level permissions. All secrets managed via Railway and Vercel environment vaults — no credentials in source code. Helmet enforces HTTP security headers. Rate limiting protects the registration and bot endpoints from abuse. Input validation applied at every API boundary.

Student data, enrollment records, and CRM lead information are not exposed in this case study. All examples use synthetic data.

Source Code & Constraints

Source code is private. The platform is in active production use. This case study describes architecture, technology decisions, and role scope. No client data, internal workflows, API credentials, or proprietary business logic is disclosed.

What I Would Improve Next

Migrate SQLite to PostgreSQL for production-grade concurrency and backup capabilities
Add a TypeScript layer across the full stack for compile-time safety
Implement a webhook-based event bus to decouple CRM state changes from Discord automation triggers
Add end-to-end observability: structured logging, error tracking, and uptime monitoring
Extend the AI bot with memory: persist conversation context across sessions for returning leads