Sandy Lim's Work | ContraWork by Sandy LimSandy Lim
Staff Data Engineer with 12 years of experience
Home Kubernetes Data Platform (3-Node K8s + Spark + Airflow + MinIO)
Built a personal lab cluster running Spark, Airflow, and MinIO on bare-metal Kubernetes, automated via Terraform and ArgoCD. Used to test Flink checkpoint tuning, Iceberg compaction strategies, and CDC pipeline patterns before productionizing at scale in enterprise environments. ML Feature Store Architecture Study (Feast + Tecton + Hopsworks)
Researched and benchmarked leading open-source and managed feature store solutions across latency, consistency, scalability, and operational complexity dimensions; published internal findings that directly informed the Project Fabricator architecture and feature platform technology selection at DoorDash. Pulsar vs. Kafka Latency Study
Built a head-to-head streaming benchmark comparing Apache Pulsar and Kafka on identical hardware, measuring end-to-end latency at 100K–500K events/sec with varying payload sizes. Published methodology and findings. Validated Pulsar's tiered storage for cost-sensitive workloads, referenced when evaluating storage options for high-volume delivery event streams.