Netflix | Software Engineer, Caching
Promoted to L4 in August 2026.
- Designed and implemented notification-driven version propagation for a Cassandra-backed publish-subscribe datastore, reducing measured publish-to-consumption latency from ~30 seconds to ~1 second; owned benchmarking and rollout with shadow validation.
- Optimized Hollow's duplicate-primary-key validator with an incremental index that checks changes between publish cycles, reducing validation time from ~28-40 seconds to ~1.5 seconds on a critical production dataset.
- Operate a fleet of ~100 EVCache/Memcache clusters, typically replicated across four AWS regions, handling deployments, scaling, instance right-sizing, and cache warming.
- Joined a second caching on-call rotation during team attrition; fixed recurring defects, deployed Memcache upgrades, and improved alerting, contributing to a ~94% reduction in monthly after-hours pages (487 to 27).
- Owned recoverable datastore deletion end to end across infrastructure, metadata, and persisted data, using idempotent workflows and soft deletes; reclaimed ~$270K in annual infrastructure costs and enabled full recovery in two production deletion incidents.
- Took over platform engineering for Hollow data-quality safeguards, adding authorized configuration APIs and partnering on a rollout reaching 437 production namespaces; enabled default validators for new namespaces and live configuration without application redeploys.
- Co-developed fast provisioning for an in-memory datastore, reducing setup from ~15 minutes to ~20 seconds; implemented namespace authorization, validation APIs, client-library extensions, and automated lifecycle tests.