Solo-Dev E-commerce & CRM Ecosystem: No-MySQL Architecture for Efficiency (VMC AIR LIFE Case Study)

Often, in the world of software development, it's common to think that building a complex platform requires teams of ten people, five-figure budgets, and massive cloud infrastructures. With the VMC AIR LIFE project, I set out to prove otherwise.

As co-founder and technical lead, I had a specific need: to create a digital ecosystem for the sale and management of VMC (Controlled Mechanical Ventilation) systems that was fast, inexpensive to maintain, and totally customizable. Off-the-shelf solutions (Shopify, WooCommerce) were too rigid for our hybrid B2B/B2C logic; enterprise solutions were out of budget.

The answer? A "One-Man Army" approach powered by Artificial Intelligence and based on a counter-current architecture. Here’s how I built an e-commerce platform with 1000+ products, a proprietary CRM, and an Android App, by myself, in fewer than 140 commits.

The Challenge: Building a Complete Ecosystem, Not Just a Shopping Cart

The requirement was not trivial. It wasn't just about "selling filters." The system needed to manage:

• Technical Catalog: Over 20 brands, cross-compatibilities, and territorial protection for brands like IRSAP.
• Management System (CRM): Customer master data, maintenance scheduling, technical intervention management, and PDF quotes.
• Italian Fiscal Compliance: Direct integration with Fatture in Cloud and SDI management.
• Performance: Zero latency, even on mobile.

The Architecture: Why I Said "No" to MySQL

The most controversial choice was the database. The industry standard is MySQL (or PostgreSQL). I chose not to use it.

Why? To eliminate overhead.

99% of e-commerce traffic is read-heavy (users browsing products), not write-heavy. Querying a remote relational database for every visit is a waste of resources for a medium-sized catalog (under 10,000 items).

I opted for a hybrid JSON + SQLite architecture:

• Frontend (JSON Flat-File): The entire product catalog resides in a single optimized JSON file (around 700KB). This is loaded into the server's RAM via APCu Cache. The result? When a customer searches for a filter, the server doesn't have to perform queries: the data is already in memory. Response time: < 50ms.
• Backend Management (SQLite): For data requiring relational integrity (customers, orders, invoices, quotes), I use SQLite. It's a file-based, extremely fast database that lives "next to" the PHP code, eliminating network latency.

The result is a platform that "flies" on standard shared hosting, with negligible running costs (~€200/year for infrastructure).

The "Pragmatic" Technology Stack

In a world chasing the latest JavaScript framework, I chose stability and robustness:

• Backend: PHP 8.4 (Modern, typed, fast).
• Frontend: HTML5 + Tailwind CSS (No complex build steps, clean design).
• Mobile: Native Android App (Kotlin) serving as an intelligent wrapper for the shop, managing Push notifications and offline cache.
• AI Integration: Google Gemini 2.5 Flash via API.

The Role of AI: My "Senior Partner"

How does a single engineer write code, manage the server, and populate a database of 1000 products? This is where AI comes in.

I didn't use AI to "write the code for me," but to amplify productivity. Integrating Google Gemini APIs into the administration panel allowed me to:

• Automatically generate SEO-friendly descriptions for hundreds of technical products from raw data.
• Analyze customer data to suggest marketing actions (e.g., "This customer hasn't changed filters in 12 months").

AI transformed weeks of manual work into processes of just a few minutes.

Beyond Code: Solving Real Business Problems

Technology is useless if it doesn't solve business problems. Two examples of how this architecture supports the company:

Dynamic Territorial Protection

As a Technical Assistance Center (CAT) for Irsap and Brofer, we have exclusivity in certain areas. I implemented a "logical geofencing" system: some spare parts are visible and purchasable only if the customer enters a unique code linked to their geographical area. A standard CMS would have required expensive and unstable plugins; here, it's native logic.

Real-time Air Quality

The customer's reserved area connects to the OpenWeatherMap APIs. If the customer lives in an area with high PM10 (e.g., Po Valley in winter), the system automatically recalculates the recommended filter replacement frequency. This is a technical added value that becomes a customer retention tool.

Conclusion: Efficiency and Innovation

The VMC AIR LIFE project (version 4.0) is proof that a solid engineering approach beats unnecessary complexity.

We achieved an Enterprise-Grade system (automatic invoicing, CRM, App, AI) with the agility of a static website.

For those interested in delving deeper into the technical aspects or similar solutions for the digitalization of industrial processes, I am available for a discussion.