Executive Summary
total delivery time
full team size
independent microservices
of code: Java, Dart, Python
public REST API, role-based access
CI/CD with vulnerability scanning
The Project
Follow.me is a location-based social app: people check in at venues, see who else is there, get acquainted and chat. The product combines a map, a social graph, messaging and identity verification, and has to remain fast and safe for a consumer audience.
The work covered the whole product, from the mobile app to the cloud infrastructure: architecture, backend services, the iOS and Android app, CI/CD and operations.
What Was Built
Mobile app — Flutter, for iOS and Android: sign-up by phone with a one-time code, step-by-step onboarding, venue check-in by QR code, a map of venues and people nearby, likes, contacts, a blacklist and privacy settings, photo albums, and one-to-one chat with disappearing messages. The app is locked with a PIN or biometrics and blocks screenshots.
- Identity verification — the user makes a gesture in front of the camera, and a separate service checks it with computer vision models (MediaPipe, InsightFace)
- 17 microservices: sixteen on Java and Spring Boot and one in Python for image analysis
- Public API of 96 endpoints with role-based access control
- Data stores matched to the task: PostgreSQL for profiles and activity, Neo4j for the social graph, PostGIS and Redis for maps and location search
- Messaging on the Matrix protocol; push notifications through Firebase and Huawei Mobile Services
Architecture
The backend follows a strict three-tier contract. Public facades hold no data and only orchestrate; business services each own their database and publish events to Apache Kafka; integration adapters wrap external products (maps, SMS, push notifications, chat, file storage) and never depend on the layers above them. Services share only typed API modules, never their internal data models.
Every request carries a correlation identifier through all services, every call to a dependency has a timeout, and database schemas evolve through versioned migrations.
Infrastructure — Kubernetes in the cloud, described in Terraform; 21 Helm charts deployed by Argo CD, with blue-green releases through Argo Rollouts and automatic scaling; 34 GitLab CI/CD pipelines that build container images and scan them for vulnerabilities with Trivy.
The Method
A team of three engineers working with AI assistants. The architecture contract, the coding policy and a reference card for every service were written first and kept in the repositories alongside the code, so that people and AI agents worked from the same rules. Features of the mobile app were developed from written specifications.
The platform was delivered in three months.
Tech Stack
Conclusions
A consumer social product is demanding in its own way: many features visible to the user, real-time interaction, location data and personal data that must be protected.
Three engineers. Three months. A mobile app for iOS and Android, 17 microservices, a public API of 96 endpoints, a social graph, maps, chat and identity verification, deployed to Kubernetes through 34 CI/CD pipelines.
The written architecture contract and per-service documentation make the platform maintainable and let another team take it over with little time spent on learning the system.