Location-Based Social Platform

Overview

A production-grade, location-based social platform for a mobile-first audience, combining a short-video feed, real-time messaging, and a deep gamification layer. The hard part was less any single feature than making all of them coexist in one app — at production quality, on mid-range hardware, built and maintained by one person.

As the technical co-founder and sole developer, I designed and built the entire product end to end — frontend, backend, infrastructure, and everything in between. The core product concept is under NDA; this case study focuses on the engineering.

Problem

Building a modern social app means shipping several genuinely hard systems at once — a performant video feed, a tamper-resistant reward economy, location features you can trust, and real-time messaging — each of which is effectively a product in its own right. Delivering all of it solo, at production quality, meant treating architecture and discipline as first-class concerns from day one. (The product thesis and go-to-market are under NDA.)

Approach

As the technical co-founder and sole developer, I own the full stack end to end: the React Native client, the Firebase backend and 109 Cloud Functions, the data model and security rules, CI/CD, the design system, and bilingual internationalization. A few of the genuinely hard problems:

A memory-safe vertical video feed

A TikTok-style infinite video feed is a notorious memory sink in React Native. I built a single-active-player model capped at three concurrent players under LRU eviction, with disciplined teardown — player release, animation stops, listener removal — on every unmount. The result is a smooth feed that stays stable on mid-range Android without leaks or crashes.

A server-authoritative reward economy

The reward economy runs entirely server-side across 109 deployed Cloud Functions, so progression can't be tampered with from the client. It's backed by a query-optimized Firestore data model (95 composite indexes) and security rules generated from consistent service patterns, for a locked-down client/server boundary.

Trustworthy location and a performant map

An interactive Mapbox map renders location-based content, backed by geohash spatial queries for efficient proximity lookups. Location-sensitive actions are validated server-side so they can be trusted, and map styling uses deterministic expressions so rendering stays stable and cacheable.

Real-time messaging

A real-time messaging system spanning direct messages, group chats, and larger communities — built on a Firestore + Realtime Database hybrid to match each pattern's read/write profile.

Result

A solo developer delivered a product with the scope and architectural complexity of a funded startup. It demonstrates end-to-end mobile engineering across real-time systems, video performance, geolocation, a server-authoritative economy, complex client state, internationalization, and native CI/CD — held together by strict architectural conventions (modular service facades instead of god objects, centralized design tokens, enforced Firebase cleanup patterns) and quality gates (automated test suites, Sentry monitoring, pre-commit linting, and dead-code scanning).

By the numbers

• 1,000+ commits over ~10 months of sustained solo development
• ~100 screens via file-based routing
• 109 Cloud Functions deployed (67 callable, 10 Firestore triggers, 9 scheduled, plus messaging and notification handlers)
• 95 Firestore composite indexes in a query-optimized data model
• 29 service domains / 200+ modules in a facade-based backend architecture
• Automated test suites across unit, integration, and service layers — coverage actively expanding
• Fully bilingual (EN/ES) across a 15-namespace i18n architecture

Status

In active development and pre-launch (August 2025 – present); not publicly listed. The product name, core concept, and screenshots are under NDA — a guided walkthrough and visuals are available on request.