TipTop CA

TipTop CA is a production Learning Management System (LMS) and EdTech platform engineered to handle end-to-end educational operations for professional Chartered Accountancy students. Serving a rapidly scaling student body of 1,000+ active learners, the platform operates not merely as a content repository, but as a full-scale transactional, identity-driven, and operational business engine.

My role spanned the entire application lifecycle as Lead Full-Stack Developer. I architected the core request lifecycle: translating responsive user interactions in Next.js into authenticated REST API commands, enforcing role-based permissions (RBAC) across Admins, Instructors, and Students, executing transactional business logic, and orchestrating synchronization across databases and external enterprise ecosystems.

At the center of TipTop CA is a modular course engine designed to deliver complex, multi-week academic syllabi with clarity. The platform supports tiered course catalogs, subject-specific lecture series, downloadable reference materials, and interactive assessment modules.

Rather than relying on heavy off-the-shelf LMS plugins that introduce latency and vendor lock-in, the course management system was custom-engineered to allow administrators to dynamically structure content, update curriculum schedules, and gate learning milestones based on enrollment status and student progress.

The student dashboard serves as the central command center for each enrolled learner. Because accountancy qualifications require hundreds of study hours, the interface was designed with minimal cognitive friction, providing instant visibility into module completion percentages, upcoming deadlines, and direct access to active lecture content.

Every student interaction—from video completion to quiz submissions—updates persistent application state in real time. Upon meeting completion criteria, the system triggers automated digital certificate generation, removing manual administrative verification overhead.

Handling educational enrollments required a resilient transactional payment infrastructure. I integrated PayFast to support diverse commercial models: one-time upfront payments, multi-month installment schedules, and high-conversion course upsell checkout funnels.

Rather than relying on the client's browser to signal payment success—a common source of failed access grants and payment disputes—the payment architecture is strictly webhook-driven. Asynchronous PayFast webhooks are cryptographically verified by the backend, updating transaction ledgers, triggering course access, and re-attempting failed calls with idempotent retry logic.

A major operational differentiator for TipTop CA was its deep bridge into the client's Microsoft and Azure enterprise ecosystem. Historically, onboarding a student into corporate educational environments required manual IT staff provisioning, creating a 24-to-48-hour delay.

I integrated Microsoft Graph and Azure Active Directory APIs directly into the post-payment webhook pipeline. The instant a student completes their initial installment or payment, the backend automatically provisions their Azure student account, configures role permissions, and provisions access credentials without human intervention.

To empower students with instant, 24/7 academic support, the LMS was upgraded with an intelligent AI chatbot. Rather than offering a generic, unrestricted conversational model prone to hallucinations, the system was built using a Retrieval-Augmented Generation (RAG) architecture powered by Google Gemini and Gemini File Search.

Course materials, lecture notes, and syllabus documentation are indexed into structured knowledge repositories. When a student poses an inquiry, the retrieval pipeline queries the specific course knowledge base and grounds Gemini's response strictly within that syllabus, preventing cross-course data leakage and enforcing student enrollment authorization.

One of the most defining aspects of this project occurred when the client significantly expanded the original codebase using AI-assisted generation tools. While features were added rapidly, it introduced critical vulnerabilities: duplicate business logic, brittle API assumptions, unhandled edge cases, and regressions across the authentication flow.

I conducted comprehensive codebase archaeology—tracing execution flows end-to-end from the React UI down through the API routers, authentication middleware, and database layers. By identifying root causes rather than applying surface-level patches, I refactored the expanded codebase, eliminated regression bugs, and reinforced authentication boundaries without disrupting live production users.

Operating a production LMS with 1,000+ students requires robust infrastructure beyond basic cloud hobby tiers. I configured and deployed the application on a dedicated Hostinger VPS running Linux, configured an Nginx reverse proxy with SSL termination, and managed application processes with PM2 for automated restarts and load monitoring.

Traffic routing, DDoS protection, and DNS are routed through Cloudflare, providing edge caching and security headers. A custom Git/GitHub CI/CD deployment pipeline was established, allowing incremental updates and bug fixes to be deployed to production safely and reliably.