Kaiser
Permanente
How virtual care could live inside the existing platform — before “telemed” was a thing anyone was building.
Kaiser already had My Doctor Online, a platform organized around individual doctors, each with their own page. Telemed wasn’t part of it and barely existed at the company. I was brought in on a genuine skunkworks effort — no KPIs, no roadmap slot, no promise anything would ship — to answer “what if?”
A per-doctor flow is a maintenance burden that grows with the roster.
If telemed lived embedded on every doctor’s page, every flow would need building and syncing across a massive physician roster — the kind of structural cost that quietly compounds with the size of the organization. The interesting design problem wasn’t the screens; it was the architecture underneath them.
One shared flow, contextual launch buttons.
The intuitive answer was to embed telemed on each doctor’s page. The spine of the concept was inverting that: instead of a telemed flow per doctor, one single shared telemed flow, with contextual launch buttons living on doctor pages. The shared home carries a persistent “Visit a Kaiser Permanente Doctor via Video” action — and renders context-aware.
It asks you to pick one
Land on the shared home with no health center and it prompts you to choose — the empty state that opens the flow.
It pre-fills everything
Arrive with one attached — a real pilot framing, the NV Health Center on the NVIDIA campus — and it pre-fills the location, address, and hours. Same page, two states, driven entirely by the context you came in with. No per-physician rebuild.
the shared home — one page, two context states
A verified wizard that forks on the same context logic.
A verified booking wizard runs a short identity gate, then forks on exactly the context logic above: the context-passed path already knows the physician; the direct path asks the patient to choose one.
the booking wizard — same steps, forks on context
Launch from anywhere; the same flow catches you.
The doctors directory (in the hero) is where the launch buttons actually live — one per physician, beside their bio and hours. And an Hours & Appointments / FAQ layer handles the questions a new virtual-care patient predictably asks, so the concept reads end to end — all feeding the one shared flow rather than a maze of per-doctor variants.
No confirmed stack — so I inferred a provisional one.
The honest hard part was that a concept exploration has no floor under it — no confirmed technology. So I inferred a provisional stack from My Doctor Online to keep moving, rather than let the uncertainty stall the design. When there are no parameters, part of the job is manufacturing your own.
I do my sharpest work inside constraints — so here I had to build my own.
The output was a coherent telemed concept whose architecture solved the real scaling problem: one shared, context-aware flow reachable from anywhere, instead of a per-doctor burden that grew with the roster. Read today, when telemed is table stakes, the point isn’t the idea — it’s what the idea is made of. The work wasn’t the screens; it was the structure underneath them.
A concept exploration with no KPIs by design — so there are no measured outcomes to quote.