Custom CMS & CRM Engine

The Problem

The client, a mid-sized e-commerce operation, had outgrown their off-the-shelf CRM system. Data synchronization was taking too long, database deadlocks were causing sporadic 502 errors during peak load, and feature development had ground to a halt due to crippling technical debt. The business required an alternative that was fast, strongly typed, and entirely owned by them.

The Solution

I designed and implemented a bespoke CMS and CRM engine from scratch using .NET 8 and Clean Architecture principles. Rather than relying on heavyweight ORMs for everything, I utilized Dapper for high-performance read projections while keeping Entity Framework Core strictly for bounded write operations.

Here is a high-level view of the decoupled architecture:

graph TD;
    Client((Web Client)) --> API[API Gateway]
    API --> ReadPath{CQRS: Query}
    API --> WritePath{CQRS: Command}
    
    ReadPath --> Redis[(Redis Cache)]
    Redis --> Dapper[Dapper Micro-ORM]
    Dapper --> DB[(PostgreSQL Read Replica)]
    
    WritePath --> Domain[Domain Layer]
    Domain --> EF[EF Core]
    EF --> MasterDB[(PostgreSQL Primary)]
    
    MasterDB -- WAL Replication --> DB

Core Implementation Details

The system heavily leverages the Mediator pattern to decouple the HTTP transport layer from the business logic.

[ApiController]
[Route("api/customers")]
public class CustomerController : ControllerBase
{
    private readonly IMediator _mediator;

    public CustomerController(IMediator mediator) => _mediator = mediator;

    [HttpPost]
    public async Task<IActionResult> CreateCustomer([FromBody] CreateCustomerCommand command)
    {
        var result = await _mediator.Send(command);
        return result.IsSuccess ? Ok(result.Value) : BadRequest(result.Errors);
    }
}

By isolating the domain, we achieved 100% unit-test coverage on the business rules.

Results

• Performance: Time To First Byte (TTFB) decreased from 1.2s to 85ms on the 95th percentile.
• Reliability: Extinguished all database deadlocks by separating read and write pathways.
• Velocity: The client’s internal team can now implement requested features in days instead of weeks.

The system is now processing over 50,000 transactions daily seamlessly.