Caden Cheah

About

Backend engineer who built a serverless AWS Lambda microservices backend for a parenting assistance mobile app, including a personalized recommendation system optimized to sub-500ms via precomputed scoring and DynamoDB caching. Demonstrates strong production pragmatism: CloudWatch-driven performance tuning (provisioned concurrency), zero-downtime phased schema migrations, and robustness patterns like optimistic locking and request deduplication. Also led a refactor of an LLM RAG pipeline to improve retrieval quality and cut latency from ~5s to ~3s.

Experience

Education

Key Strengths

• Designed end-to-end serverless backend for a startup mobile app (AWS Lambda microservices)
• Built hybrid recommendation architecture with precomputed scores + DynamoDB caching achieving sub-500ms responses
• Used CloudWatch metrics to target high-impact performance work (auth/activity/recommendation endpoints)
• Reduced latency ~40% by reusing DynamoDB client connections across Lambda invocations
• API reliability via Pydantic validation and versioned schemas with backward compatibility
• Zero-downtime, phased database migrations (nullable columns → dual-format code → backfill → tighten constraints)
• Refactored LLM RAG pipeline improving response time from ~5s to ~3s with read-only fallback for risk mitigation
• Improved robustness with optimistic locking (DynamoDB conditional writes), deduplication, and short-lived per-user request cache
• Implemented JWT auth via Firebase plus audit logging (user/action/timestamp)