Problems that looked operational. Systems that had to change.

Generator Health AI

Pressure-testing AI before the workflow depended on it

Validated five prior-authorization AI applications against real operating conditions: payer rules, production cases, usability failures, false positives, false negatives, and frontline edge cases. The job was not to prove the product worked. It was to find where it didn’t — before users had to discover that themselves.

Generator Health AI

Turning denial prediction into an operational decision

A risk score only matters if someone knows what to do with it. Worked across the workflow around predictive denial signals to define how flagged cases should change staff behavior before submission — moving the value of the model upstream from post-denial cleanup toward prevention.

Mount Sinai

Rebuilding the Epic workflow around how the clinic actually ran

Led a 12-week redesign across scheduling, referrals, documentation, care-team communication, and clinic coordination. The work connected Epic configuration to the reality of a high-volume addiction medicine program rather than treating the EHR as the workflow itself.

Amedisys

Finding the operational leak behind the revenue loss

Used Power BI and operating data to trace where value was being lost across four departments, reconcile conflicting reporting, and focus recovery work on the highest-value failure points instead of treating the financial outcome as an isolated problem.

Mount Sinai

Designing capacity around the actual day, not the ideal one

Redesigned daily clinic operations across provider schedules, room use, patient flow, same-day access, behavioral-health escalation, and team communication. The goal was not a cleaner process map; it was a clinic that could absorb real variability without losing control of the day.

Mount Sinai

Building the operating system around long-acting injectable access

Built an end-to-end workflow spanning medication access, ordering, inventory, scheduling, administration, forecasting, and Epic documentation. The intervention connected fragmented steps into one operating lifecycle instead of leaving staff to manage the gaps manually.