Selected systems I have built across AI tooling, retrieval infrastructure, and fulfillment operations workflows.
An internal system I architected to combine webhook-triggered enrichment, platform and shipment context, retrieval across internal and public knowledge bases, and bounded AI actions inside one support workflow.
Reduced support investigation time by an estimated 70-80% and removed constant context switching between systems.
Built a customer-facing support chatbot with chunk-based semantic search, conversation memory, Zendesk content sync, and user-type filtering for different merchant audiences.
Estimated 32% reduction in inbound support tickets, supported by a feedback system designed to improve context quality, intent handling, and retrieval performance over time.
Used AI and historical Zendesk ticket data to identify weak spots in the knowledge base, define a more structured article format, and generate improved or net-new help content for retrieval workflows.
The goal was not just better articles. It was a better knowledge system for AI-assisted support.
Built a game inspired by the viral cup swap trend. Drag colored cups into the right sequence before the timer runs out. Has easy and hard modes, a global leaderboard, and I shipped it as an iOS app too.
A side project exploring structured memory, durable context, and how useful systems can preserve the right information over time.