Let's go CRM
Customer service and driver CRM via WhatsApp for a Venezuelan ride-sharing platform, with human agents and an AI sharing the same conversation inbox, under hexagonal architecture + Feature-Sliced Design.
The Challenge
WhatsApp support depended on multiple external data sources that the company does not control (Meta API, proprietary driver/trip APIs, the AI model). The system needed to isolate its business logic from those sources so it wouldn't break with every external change, and respond in real time without losing messages under load.
The Solution
A CRM where human agents and an artificial intelligence share the same WhatsApp inbox: if no one is attending a conversation, the AI responds on its own, and as soon as an agent takes it over, the client doesn't even notice the change. That AI doesn't stay idle: it learns directly from the corrections made by the team itself, so each corrected response makes it a bit more accurate, without retraining anything manually. All of this runs without losing a single message even if many arrive at once, and it stays standing even if something changes on WhatsApp's side, driver data, or the AI provider.
Focused on
Delivery and ride-sharing companies with WhatsApp support
Role in the project
Solutions Architect and Full Stack Dev.