Victor Siu

Built the scalable data and AI foundation for HubSpot’s CRM and Breeze AI layer, including multi-tenant isolation, real-time event processing, and AI experiences for 300k+ customers. Owned backend services supporting object/record/association persistence, AI orchestration, and workflow grounding. Developed microservices in Java, Python, Node.js, and Kotlin using Vitess/MySQL and HBase, including email/domain deduplication, property update paths, and eventual consistency for high-volume imports and complex filters. Implemented RAG pipelines and CRM-integrated context/tool orchestration for Breeze Assistant and agents using Qdrant (20B+ vectors), grounded responses via OpenAI/Anthropic, and built Kafka stream topology for object update events and workflow rollouts with priority lane isolation. Led multi-region AWS infrastructure on Kubernetes, configuring sharding, Elasticsearch indexing, Redis caching, and Kafka clusters across regions to enforce data residency and horizontal scale. C

Owned the CRM data platform and Breeze AI foundation at multi-tenant scale, including object/record/association services and AI orchestration grounded in customer data. Built Java/Python/Node.js/Kotlin microservices for persistence and bidirectional associations across data stores, implemented RAG pipelines backed by a vector database with 20B+ vectors, and designed Kafka topic routing/swimlanes for real-time event processing with low latency. Delivered React/TypeScript interfaces for CRM records and AI chat/agent configuration surfaces, and led multi-region AWS infrastructure on Kubernetes (sharding, indexing, caching, Kafka clusters) to enforce isolation, data residency, and horizontal scaling. Contributed to CI/CD and operational monitoring, added health checks and guardrails, and mentored 5+ engineers on multi-tenant patterns, Kafka strategies, and AI grounding.

Engineered core ticketing, real-time handoff, and telephony systems on Zendesk’s high-scale Rails platform, supporting tens of millions of daily tickets and calls while extending automation, AI suggestions, and API reliability for global support workflows. Implemented ticket creation, status transitions, and audit logging in a sharded Ruby on Rails monolith backed by sharded MySQL, with routing and replication-safe ads that kept status updates consistent across 200+ shards while handling 10M+ tickets per day. Extended Rails business-rule engine with new condition evaluators and action handlers for triggers and hourly automations, processing ~50,000 rule evaluations per minute and reducing routing re-fire loops that previously delayed high-volume accounts. Built chat-to-ticket linkage and agent handoff paths using GraphQL Conversations API over websockets, supporting 15,000 concurrent sessions and reducing handoff latency by ~40%. Integrated Twilio PSTN/SIP via Ruby services for IVR t