Beeasist — SaaS CRM and AI assistant
A high-traffic SaaS CRM platform for appointment-based businesses, where I ran the architectural design, the AI layer and both backend and frontend.
- Laravel
- PHP
- PostgreSQL
- Redis
- Next.js
- React
- Tailwind CSS
- MCP
The product
Beeasist is a SaaS CRM platform for appointment-based businesses. Multi-branch structures, staff calendars and dynamic booking rules sit at the centre of the product — which makes concurrency and the calendar model its most critical engineering problem.
I ran the platform’s architectural design, its AI layer, and both the backend and frontend.
The AI layer
I engineered a custom MCP (Model Context Protocol) infrastructure on Laravel. Integrating Prisma PHP and the Vercel AI SDK, I deployed an AI assistant that lets users drive operational workflows in natural language.
The real work here is not wiring up a model; it is defining the boundaries of the tools the assistant can reach. An assistant that can create a booking should not, by the same token, be able to delete one by accident.
Booking and concurrency
I developed a highly concurrent appointment optimisation architecture supporting multi-branch setups, staff and dynamic calendar management. A correctness guarantee that does not leak between two simultaneous requests for the same resource is not negotiable in a product like this.
Infrastructure and performance
- Optimised the Laravel backend with PostgreSQL and Redis queues to absorb traffic spikes.
- Moved push, SMS, email and WhatsApp notifications onto async processing, so delivery latency no longer shows up in request time.
- Integrated Sentry for production monitoring, so bottlenecks and errors surface before a user reports them.
Application layer
I built financial tracking and reporting modules, plus online payment integrations, from scratch over REST APIs. Web and mobile apps are built with Next.js, React and Tailwind CSS, with state centralised in Redux Toolkit so both clients share the same business rules.