The work that happens before a single polygon exists.
A VR build lives or dies on decisions made months before anyone puts on a headset. My job was the paper layer — turning "we should teach anatomy in VR" into something an engineering team could actually build, on a budget leadership could actually approve.
Which simulations ship first — and what kind of thing each one is
Three simulations made the cut: Basic Wound Dressing, Hand Hygiene & PPE Protocol, and Human Anatomy. The critical scoping call: the first two are step-by-step procedures, Anatomy is an open exploration — two fundamentally different interaction models, scoped as such from day one instead of forcing everything into one template.
Turning learning goals into buildable specs
The BRD translated "students should understand anatomy spatially" into concrete requirements: 8 toggleable anatomical systems, 3 interaction modes, male and female models, grab-and-examine with snap-back reassembly, and full onboarding — safety instructions and controller mapping — before first entry.
What it costs — and where the leverage is
Costed the build per simulation and weighed the delivery economics: a headset-only experience means every learner needs hardware. The web access path became the cost lever — the same simulations, running in a browser inside the Emverse app, meant the platform could scale to every enrolled student without scaling the headset fleet.
Sequencing the build so it actually shipped
Structured the delivery timeline across the three simulations and both platforms — sequencing what shipped when, keeping onboarding and safety flows inside the first release rather than a fast-follow, and holding the line on VR-web parity so neither path launched as a second-class experience.
Two doors into the same room.
The single most consequential scoping decision: don't gate learning on hardware. Every simulation is reachable through the headset and through the browser — same catalogue, same experience, same anatomy.
A headset per student doesn't scale. A browser tab does — and the learning shouldn't wait for the hardware budget.
