GameDuo

Resolved delayed balance issue response caused by inability to track data change history during game operations

• Git-like Version Control System: UUID-based cross-environment/project entity tracking ensuring data consistency across 6 games
• Event Sourcing-based Change Tracking: Auditable decorator + TypeORM Subscriber pattern for automated entity change recording pipeline
• 3-Way Merge Engine: Multi-environment version merge system with parent/child entity conflict detection and unique constraint handling across 6 games
• Version Diff Engine: Dual strategy of incremental comparison and snapshot comparison based on Base Audit availability for cost-efficient version diffing
• Shared identifier-based Identification: Decoupled from PK dependency, ensuring accurate entity tracking during migration, comparison, and merge

Unified marketing platform managing Google Ads/Meta/TikTok campaigns, creatives, and metrics in a single system, built on dedicated marketing DB separation

• Unified Campaign Automation: Built API for managing Google Ads/Meta/TikTok campaign creation, deployment, and modification in a single platform
• Creative Auto-deployment Pipeline: Integrated pipeline from creative upload to automatic registration on Google/Meta/TikTok platforms
• NAS-S3 Dual Sync: SQS event-based server-worker separation architecture for bidirectional sync/restore between internal NAS and S3
• Meta Asset Sync Performance Optimization: ORM single-row saves to bulk operations conversion reducing Image sync time by 72% (25.7s→7.1s), DB transactions by 95% (10~16s→0.3~0.5s)
• Lambda ARM64 Migration: Graviton2 architecture migration reducing creative processing Lambda costs by 20%
• BigQuery Storage Read API Adoption: gRPC streaming-based high-performance data reading reducing BigQuery costs by 82% ($6.25→$1.1 per TB)
• DB Optimization: Normalized 100+ column table (main/time-series/prediction split), index optimization and cursor-based pagination improving query performance from 18s to 0.5s
• Retention Metrics Integration: Consistently applied retention metrics across the full pipeline from BigQuery ingestion to entity, API, and field metadata for user retention analysis

Resolved batch job limitations from marketing metrics 60-day → 360-day expansion and serverless transition for batch processing

• Hybrid Architecture: Maintained API server (EC2) + separated batch/job processing (Lambda) for optimal resource utilization per workload
• Event-Driven Architecture: SQS + Lambda + EventBridge-based async processing reducing batch processing time from 2hrs to 5min, expanding data collection range 6x (60 days → 360 days)
• Transactional Outbox Pattern: Scheduled and delayed event publishing ensuring data consistency
• Lambda Operational Stabilization: Batch Size bulk processing resolving throttling, CloudWatch log optimization reducing costs, build OOM remediation
• RDS Proxy Connection Management: Resolved DB connection exhaustion during massive Lambda execution via RDS Proxy pooling

Built S3-based sync and automated DDL management system to resolve dynamic game data inconsistency across environments

• Cross-environment Data Sync: S3-based dynamic game data synchronization across development/staging/production environments
• Automated DDL Management Engine: Dynamic PK column type resolution, column type mismatch detection with MODIFY, and automatic index creation/RENAME for schema sync automation
• Save/Upload Performance Improvement: Analyzed and optimized Cloud Data ingestion and S3 upload bottlenecks, resolving large-scale processing delays
• Sync Stabilization: Job separation, scheduling approach transition, timeout adjustments, and non-blocking processing for operational reliability

Development and operation of NestJS utility (10 modules) + Game Server Kit (2 packages) for multi-project code consistency

• 10-module Library System: Expanded to core, repository, cache, lock, slack, crypto, smb, hash, type, and iac modules
• Repository Module: Bulk operations, Audit Log, TypeORM type narrowing support; refactored 2000+ line codebase into SRP-based file separation
• Distributed Lock Module: ElastiCache (Redis)-based distributed lock decorator for concurrency control in multi-instance environments
• TypeScript Generic-based Type Safety: Overcame TypeORM type limitations through function overloading and generics
• Internal NPM Package Management: GitHub Packages-based distribution with CI optimization testing only changed packages
• Game Server Kit Package Extraction: Extracted Sheet module + 5 submodules from game server monolith into @gameduo/glider-sheet package for multi-game project reuse, with 5-branch divergence analysis and conflict resolution

Built probability calculation package and CDK-based audit log analytics infrastructure to address regulatory risk from lack of game probability verification

• NestJS DynamicModule-based Probability Package: Packaged 5 probability functions + automatic Kinesis logging as @gameduo/glider-probability, integrated across 12 game projects
• CDK-based Analytics Pipeline: Codified Kinesis → Firehose (Dynamic Partitioning + Parquet conversion) → S3 → Glue → Athena end-to-end probability audit log analytics infrastructure
• LocalStack Integration Testing: Replaced mock tests with Testcontainers-based integration tests, achieving 12 suites 115 tests with 94%+ coverage

Games on AWS 2025 Customer Session

Shared case study of a junior developer scaling data pipeline capacity 270x in just 10 days