Selected Work

Case Study

Multi-channel customer support architecture with LLM-assisted triage

Designed team-based ticket routing, retention flows, and the integration of an LLM library for understanding customer conversations.

Customer SupportLLMArchitectureFintech

Context

A fintech company's support team handled inbound tickets through a shared inbox. Volume was growing faster than the team, response times were slipping, routing was inconsistent, and agents context-switched constantly between unrelated cases.

The team needed structured routing, escalation rules for account-closure intent and dispute scenarios, and a way to use LLMs without rebuilding the conversation layer.

Approach

I designed team-based ticket routing on a major customer-support platform — cases reach the right team based on customer attributes and content, not random round-robin. Retention and escalation logic now intercepts high-signal cases (closure intent, disputes) before they fall through.

I also migrated the web cancel-flow to integrate directly with the support platform so retention attempts happen in-product, not over email. The LLM integration sits next to the routing layer: it extracts intent, sentiment, and urgency, and feeds those signals back into routing and SLA decisions.

My role

Owned the end-to-end architecture. Partnered with support leadership on the workflow design — which queues, which escalation rules, which retention triggers — and was the DRI for the LLM integration.

Outcome

Tickets now route automatically to the right team. The cancel flow integrates with retention attempts before churn, instead of after. And the LLM signals laid the groundwork for further AI-assisted support workflows the team is building on top.