Zeta | Devpost

Inspiration

The idea started with a straight-talk conversation with Jake Levy, COO of ZaiNar, ahead of the UW Spatial Intelligence Ideathon. He described a scene almost in passing: a hospital room turns over between patients through a chain of handoffs — nurse in, doctor in, nurse out, cleaning crew in — coordinated today by phone calls and whiteboards. That one detail sent us digging into hospital operations research, and we kept running into the same number: nurses spend less than 20% of a shift on direct patient care. The rest goes to locating people, locating equipment, and managing handoffs that a computer could be tracking instead.

The sharpest version of that problem turned out to be patient falls and elopement. Roughly 700,000 to a million hospitalized patients fall every year in the US, a third of those falls cause injury, and CMS treats a fall with injury as a "never event" — meaning Medicare pays nothing toward the extra care it causes. Hospitals already know this is expensive. What they don't have is a way to watch every at-risk patient continuously without either paying for a human sitter or putting a camera in a room where cameras are flatly not allowed. That gap — a monitoring problem cameras structurally cannot solve — is where Zeta came from.

What it does

Zeta is a camera-free, through-wall patient safety layer for fall- and elopement-risk patients, built on two technologies most people wouldn't think to combine: ZaiNar's RF-based indoor positioning and SGA-AR's spatial mapping.

Each patient at risk wears a small tagged wristband. ZaiNar tracks that tag continuously, through walls and around corners, to sub-meter accuracy — no line of sight, no camera required, which matters because patient rooms and behavioral health units are places cameras are not allowed to go. Separately, a one-time SGA-AR spatial scan of the unit builds a zone map — bed, bathroom, hallway, exit — so raw coordinates become meaningful locations. When a patient crosses a boundary they shouldn't — heading to the bathroom alone at 3 a.m., or approaching a stairwell on a wandering-risk unit — the nurse gets an alert before it becomes an incident, not after.

It doesn't replace nursing rounds; it changes what they're for. Rounds still happen on whatever cadence hospital policy requires, but most in-room checks become quick confirmations instead of blind searches, because the system already knows where every at-risk patient is.

How we built it

This was a concept-and-validation sprint, not a code sprint, so "building" meant building the argument as rigorously as we'd build the product. We started from the ideathon's own five-question framework — pick one person, name their current workaround, define what changes with shared spatial awareness, find who pays, and define the smallest first version — and refused to move to the next question until we could answer the one before it in plain language, no technology jargon.

From there we pressure-tested the idea against real numbers instead of assumptions: hospital fall incidence and injury-cost data, CMS's never-event reimbursement policy, published hourly wages for hospital sitters, the ratio used in existing video tele-sitting programs, and public pricing ranges for hospital RTLS hardware deployments. Every dollar figure in the pitch — the $480/day sitter cost, the $210/day tele-sitter cost, the $60/day Zeta target, the $25K–$40K pilot install versus $875K–$2M for hospital-wide UWB systems — is sourced, not invented, so the business model would hold up under a judge's first follow-up question.

We then built the deliverables a founder would actually need: a nine-slide pitch deck designed around the judging rubric's four categories, and a timed four-minute two-speaker script built to the deck, so the story that's read matches the story that's spoken.

Challenges we ran into

The hardest problem wasn't finding an idea — it was finding one where both technologies were genuinely load-bearing instead of one being window dressing. Several early directions (equipment tracking, room-turnover orchestration) leaned so heavily on ZaiNar that SGA-AR became decorative, which fails Track 2's actual requirement that both technologies be essential. We only found the right shape once we asked what SGA-AR could uniquely provide that ZaiNar's raw coordinates could not — a zone map that turns "patient is at X,Y" into "patient is in the bathroom" — and confined SGA-AR honestly to that one-time setup role rather than overstating it as a live component.

The second challenge was pricing something that doesn't exist yet. Public RTLS pricing data is sparse and hospital procurement is opaque, so we anchored our numbers to the closest verifiable comparisons — sitter wages, tele-sitting staffing ratios, and published hospital RTLS deployment costs — and stated our assumptions explicitly rather than presenting a guess as a fact.

Accomplishments that we're proud of

We're proud that every claim in the pitch survives a follow-up question. The cost comparison, the fall statistics, the CMS reimbursement policy, and the RTLS pricing benchmarks are all sourced from public data, not asserted from confidence. We're also proud of how cleanly the tech-fit argument closes: because cameras are categorically prohibited in patient rooms and behavioral health units, ZaiNar's non-visual RF tracking isn't the best sensing option available, it's the only one — which is a much stronger claim than most RTLS pitches can make. And we're proud that the pitch is buyer-first: nursing administration and risk management are shown a system that substitutes for money they are already spending, not a new budget ask.

What we learned

We learned that a strong hackathon concept and a strong business case are the same exercise once you take the judging rubric seriously — commercial clarity, problem quality, and tech fit all collapse into one question: can you say, specifically and without hedging, who has this problem, what they do about it today, and why they'd pay you instead. We also learned how easy it is to blur user and buyer in healthcare pitches, and how much sharper a concept gets the moment you separate "who feels the pain" from "who signs the check." Finally, we learned that citing a real number — even an imperfect, publicly sourced estimate — beats a confident-sounding guess every time a judge starts asking follow-up questions.

What's next for Zeta

The immediate next step is exactly what the ideathon brief promises to strong concepts: a real conversation with ZaiNar and the UW lab about whether this is worth building. Concretely, that means validating the pilot economics with an actual hospital's patient safety or risk management office, refining the $60/patient-day target against real sitter payroll data from a design-partner hospital, and prototyping the SGA-AR zone-mapping workflow so we can show, not just describe, how fast a unit can be set up.

Beyond the first pilot, Zeta's roadmap follows its own wedge-and-expand logic: prove the KPI on one med-surg or geriatric unit, extend to psychiatric and behavioral health units with tamper-resistant, ligature-safe tags where elopement risk is highest, then to memory care, then hospital-wide and across health systems. Longer term, the same infrastructure — precise, camera-free, through-wall location tied to a spatial zone map — is a foundation other clinical safety products could be built on, not just this one.