Gaming Platform Infrastructure
Multi-tenant event streaming platform processing 10M+ events/day with zero data loss across 15+ microservices
15+ microservices
438 event types
The Problem
Legacy analytics silos, no centralized telemetry, no multi-tenant platform for scaling across titles
Legacy analytics silos, no centralized telemetry, no multi-tenant platform for scaling across titles. A mobile gaming studio with multiple live titles relied on fragmented third-party analytics platforms that offered no raw data access, no custom instrumentation, and no path to centralized product intelligence. Each title ran its own analytics stack. Player behavior data was trapped in vendor dashboards with no export path. The studio needed a platform that could ingest, transform, and surface telemetry from any title through a single architecture.
We built a multi-tenant event streaming platform on AWS. A custom Unity SDK batches and ships game events to a Go telemetry service running on EKS with Graviton4 processors. Events are validated against a schema registry, authenticated via JWT, and published to Kafka topics — one per title — for downstream consumption.
Amazon MSK Kafka feeds 12 Firehose delivery streams that land raw events into an S3 data lake organized as a medallion architecture. A dbt pipeline running hourly on ECS Fargate transforms raw Bronze events through a Silver cleaning layer into Gold KPI models — retention cohorts, daily active users, revenue segmentation, and economy balance metrics. 30 dbt models with 136 automated data tests ensure zero drift between raw telemetry and executive dashboards.
The platform was stress-tested at 1,700 sustained events per second with zero data loss — validated against a known event count. Rate limiting, distributed tracing via OpenTelemetry, and 112 CloudWatch alarms provide production-grade observability from client to dashboard.
Event types
Sustained throughput
Automated data tests
Editorial notes
Mandate
Replace fragmented third-party analytics with a centralized, multi-tenant telemetry platform that the studio owns end-to-end.
Signal
The value was not instrumenting more events. It was building the streaming spine that lets any title plug in and produce executive-grade product intelligence from day one.
Context
A mobile gaming studio with multiple live titles relied on fragmented third-party analytics (Unity, Firebase) with no raw data access, no centralized telemetry, and no path to multi-title product intelligence.
Constraint
Each title ran its own analytics silo. Player behavior data was trapped in vendor dashboards. Custom instrumentation and cross-title analysis were impossible without a unified platform.
Intervention
Built a multi-tenant event streaming platform: Unity SDK → Go/EKS telemetry service → Kafka → S3 medallion data lake → dbt transforms → Metabase dashboards. 15+ microservices, 438 event types, 136 automated data tests.
Outcome
Platform proven with live player data at 1,700 events/sec sustained, zero data loss. 30 dbt models powering retention, revenue, and economy dashboards — delivered in 35 weeks.
Architecture
From game client to executive dashboard in a single streaming architecture
Event Ingestion
Stream Processing
Data Transform
Analytics & Observability
Tech Stack
Backend
Go 1.24+ (EKS, Graviton4/ARM64)
Streaming
Amazon MSK Kafka (27 topics)
Data Lake
S3 Medallion (Bronze / Silver / Gold)
Transform
dbt + Amazon Athena (30 models)
Compute
EKS + ECS Fargate + Firehose
Observability
OpenTelemetry + X-Ray + Prometheus
Results
Planning a Similar Mandate?
A direct working session about the problem, the constraints, and the fastest credible path forward.
We respond within 4 hours during business hours