Portfolio

Healthcare AI and system governance work

Case studies and advisory patterns for teams building or buying algorithmic systems whose decisions need to be explainable, challengeable, and correctable.

Production identity and verification systems 300K+ clinicians have used Dialer for over 110M+ video and audio visits.

Coordination tools that work when things get busy Clear response windows, escalation paths, and records of what changed so teams recover quickly under real conditions.

Product, legal, and operations alignment Turn requirements into delivery behavior so growth stays manageable in regulated settings.

Doximity Dialer caller ID verification interface on iOS
Doximity

2013–2017

Verified caller ID for HIPAA-conscious telehealth

Patients ignored calls from unknown numbers, stalling telehealth. Dialer showed the doctor's office number — verified and HIPAA-safe.

Read full case study
Key outcomes
  • Patients ignored unknown numbers, blocking telehealth adoption.
  • HIPAA privacy, carrier rules, and clinician adoption needed to work together.
  • Pickup rates improved enough that clinics stopped raising it as a top concern.
  • iOS app reviews climbed from 3.7 to 4.8 stars during mobile product work.
Transcarent care navigation app — public app store visual
Transcarent

2021–2022

Care navigation that cut avoidable escalations

Led product work across value-based specialty-care and care-navigation programs (Surgery, Urgent Care, Behavioral Health, Oncology Care).

Read full case study
Key outcomes
  • Guidance was fragmented across programs, so every case was an exception.
  • Managed product managers and senior product managers across specialty-care programs.
  • Routing and escalation rules cut the exception queue by making ownership visible.
  • Care plan completion rates converged — the gap between highest and lowest programs shrank.
Andwise physician financial planning interface
Andwise

2022–2024

Andwise: fiduciary planning & physician support

Co-founded Andwise; public pages corroborate founder and advisory-board context, while growth and funding metrics are founder-reported.

Read full case study
Key outcomes
  • Clinicians needed financial guidance without sales pressure.
  • Fiduciary duty, compliance review, and physician context had to work together.
  • Recommendation turnaround dropped from days to hours for standard reviews.
  • Grew to 1,200+ physician users and a 700-member physician community.

Everything shared stays shared only with your consent. Internal politics stay unnamed, identifying details are never published without approval, and metrics are rounded and de-identified. Redacted versions of measured changes are shared only when you choose.

Governance review

A focused review of an existing or planned AI/algorithmic system: decision rights, escalation paths, evidence quality, and correction workflows.

Written assessment + prioritized action list

Workflow design

Shape the human-AI handoff so clinicians and operators can trust, challenge, and fix recommendations without slowing care.

Workflow maps, decision trees, and ownership specs

Advisory retainers

Ongoing advisor for product, compliance, or operations teams navigating multiple AI-enabled decisions.

Monthly calls and async review

Send a few sentences about what you're working on and where you're stuck. I'll reply within 48 hours — usually with a question or two to sharpen the problem, and a sense of whether there's a fit. If there is, we'll do a 30-minute call to dig into the details. No decks, no proposals. We figure out whether there's something to work on before anyone commits to anything.

Availability

I'm open to conversations about healthtech products with compliance, safety, or coordination complexity.

Send a paragraph about what you're working on and where you're stuck.

A short call to see if there's a fit. · Reply within 48 hours.

Confidentiality

Everything shared stays shared only with your consent. Internal politics stay unnamed, identifying details are never published without approval, and metrics are rounded and de-identified. Redacted versions of measured changes are shared only when you choose.