Smovee

As a freelance developer, I was approached to build more than just a music app—it was a civic engagement project with a clear mission. The challenge was to create a digital space for refugee, displaced, and migrant artists, allowing them to showcase their stories and preserve their creative heritage, often over unstable, low-3G connections. This project pushed every technical limit and taught me that the most powerful code is driven by empathy.

The goal was to build a lifeline that could amplify their voices, build community, and enable support.

The Real-World Problems

The mission was to address a distinct set of human and technical challenges:

• Unstable Connections: We had to deliver a high-quality audio experience over unreliable, low-speed 3G networks.
• Amplifying Voices: Displaced artists needed a platform to be heard, preserving their cultural sounds and stories for a global audience.
• Building Community: Artists felt isolated and needed tools to connect with listeners and other creatives to build supportive communities.
• Enabling Support: Without access to traditional channels, artists required a way to gain both financial and social support for their work.

Crafting the Solution, One Challenge at a Time

Every technical decision was a balancing act between modern features and accessibility.

The Streaming Dilemma

Our biggest hurdle was providing premium sound over weak internet. How do you provide uninterrupted music when the connection is an unstable 3G signal?

Our solution was an adaptive HLS streaming system built with FFmpeg. Each uploaded track is automatically transcoded into multiple bitrates, allowing the player to dynamically switch quality to match the available bandwidth. This ensures a smooth listening experience even on the weakest networks.

A More Human Recommendation Engine

To create playlists tailored to unique tastes and cultures, I engineered a sophisticated hybrid recommendation system.

First, the system intelligently avoids listener fatigue by excluding tracks the user has recently played or liked, ensuring every suggestion feels fresh. From there, it runs a multi-pronged strategy:

1. Acoustic Similarity: Using PostgreSQL with pgvector, the engine analyzes the actual audio of every song to find others with a similar sonic fingerprint.
2. Collaborative Filtering: It identifies "taste twins" and recommends music popular within that group that the user hasn't discovered yet.
3. Personalized Genre Mixes: It finds tracks trending within a user's favorite genres for highly relevant mixes.
4. Artist-Driven Signals: A "Release Radar" ensures users never miss new music from artists they follow.

The system then assembles and scores dozens of conceptual playlists on relevance, diversity, and freshness. Only the best are presented, with clear titles like "More like [Song Name]" to create a transparent and personal experience.

Features That Empower

To fulfill the mission, I focused on tools for storytelling, healing, and community building:

• Artist Dashboard: A simple interface providing analytics for artists to understand their audience and track their support.
• Direct Upload & Verification: A streamlined process for artists to control their own profiles and music.
• Curated Playlists: Themed collections that celebrate specific cultural traditions, helping artists get discovered.
• Story Sharing: A dedicated space for artists to share the rich stories and context behind their music.

Project Status and Presentation

The platform was successfully developed and deployed, and is now hosted as a live demonstration for this portfolio. The project serves as a comprehensive showcase of my ability to deliver a complex, full-stack application from concept to launch.

It effectively demonstrates the core technical achievements, including the adaptive streaming pipeline, the AI-powered discovery engine, and the scalable infrastructure. The current version stands as a powerful proof-of-concept, with a solid technical foundation ready for future development.

What I Learned

This project was a profound learning experience that reshaped how I approach technology:

1. Empathy is the Ultimate Spec Sheet: The best technical decisions came from deeply understanding our users' constraints.
2. Performance Isn't a Feature; It's a Prerequisite: In constrained environments, performance is the bedrock of the user experience.
3. Build for Resilience, Not Perfection: A system that gracefully handles failure is more valuable than one that assumes a perfect environment.

Technology Stack

• Backend: Laravel 12, PHP 8.2+
• Frontend: React 18, TypeScript, Inertia.js
• Database: PostgreSQL with pgvector
• Streaming: HLS.js, FFmpeg
• Search: Meilisearch
• Cache/Queue: Redis, Laravel Horizon
• AI/ML: Audio embeddings, Collaborative filtering
• Infrastructure: Kubernetes, Docker
• Real-time: Laravel Reverb (WebSockets)
• Performance: Laravel Octane with FrankenPHP