Juan Velasquez

Owned and rebuilt the backend that drives performance benchmarking across FlashArray: Python services responsible for orchestration, scheduling decisions, and result aggregation. Inherited a system where engineers manually triggered runs, babysat retries, and had no confidence in results. Redesigned it as an event-driven, experiment-scoped workflow: declare intent, let the system handle execution, rescheduling, and failure recovery. Separately, loaned cross-team to lead infrastructure modernization for the SSU platform, the release toolchain that every Pure Storage software shipment (FA, FB, CBS, Krypton) flows through.

• Built Python backend services for orchestration, scheduling, and result aggregation at experiment scale
• Replaced a REST service-to-service API with gRPC: lower latency, ~100 fewer worker replicas at peak (observed post-deployment)
• Implemented event-driven autoscaling with KEDA on AWS EKS; migrated deprecated SimpleDB to MongoDB
• Took ownership of a ~30k LOC Python codebase, added tests to make performance results reproducible and trustworthy
• Contributed to ~$1.25M in annual cost savings through infra and workflow improvements
• Leading ~25-service migration from EC2/Spinnaker to Kubernetes, addressing 10+ open incidents and a declining SSU upgrade success rate
• On-call rotations, incident response, postmortems; technical interviewer

• Built Python/Flask backend services and REST APIs for orchestrating remote hydroelectric turbine simulations, eliminating the need for on-site engineers
• Integrated physical lab hardware with virtualized hosts; parallel simulation runs via a React interface

• Refactored data pipeline orchestration via multithreading, scheduling, and cleaner abstractions, achieving ~45% throughput improvement

• Built backend automation for transcription and subtitle generation, cutting processing time by ~40%

• Built automation tools saving ~80 work hours/month, 50% reduction in manual data entry