40K+ backlog cleared in 3 days, $110K+ in annual savings.

Problem

A health system's front-line staff were buried under a growing queue of notifications that each needed manual handling, pulling time away from patient-facing work.

What I did

Ran user interviews with the staff doing the work, scoped two AI notification-clearing automations around what they actually needed, and pitched the client CIO to secure buy-in before building.

Shipped

Led both automations from design through go-live, defining launch KPIs and success criteria up front.

Bigger picture

These sat inside a portfolio of 17 automations I owned at the account, which held 99% uptime and drove $3.8M in annual impact (6x ROI). I also rebuilt a legacy workflow that was mismatching patient data via direct API integration, raising throughput 500%; that approach was later extended to Cleveland Clinic.

Better allocation decisions across 126+ workflows, ~$60K in annual labor savings.

Problem

More than 20 engineers across time zones were split across 126+ concurrent workflows, with no single view of who had capacity and where work was piling up.

What I did

Designed a capacity-scoring model and built the dashboard tool myself using Claude Code, alongside the company's first workflow registry.

Shipped

Used it to run weekly allocation syncs with the global engineering team, turning staffing from guesswork into a shared, data-backed conversation.

Rules page usage up 15% after research-driven design changes.

Problem

Candid was rolling out a feature that let customers create their own billing rules, but early users were struggling with the rule creation flow.

What I did

Led user feedback sessions during the beta, mapped where people got stuck, and translated those findings into specific UI and UX changes with Product Design.

Related

Earlier, I unblocked a $300K ACV customer stalled for months by aligning Product and Engineering on a new approach and shipping their billing rule changes myself via GitHub, cutting their support tickets by 90%.