222
An autonomous Far-UVC sanitising ecosystem for rideshare vehicles.
Rideshare drivers absorb the cost of post-pandemic cleanliness expectations twice over: unpaid minutes between fares, and chemical wipes degrading an interior they own. 222 moves sanitisation into the time the car is already working — including the minutes a passenger is sitting in it.
- 222nm
- Occupant-safe wavelengthWithin ACGIH limits
- 99.9%
- Deep Clean targetBoth modules, modelled ~25 min
- 0
- Minutes of manual wipingThe point of the system
- Role
- Hardware design · App UX · Research
- Team
- Ultimanium — SCAD IACT 330
- Year
- Concept project
- Recognition
- 5 × Indigo Design Awards
Drivers were paying for cleanliness with their own time.
We interviewed rideshare drivers at SCAD, and the complaint that kept arriving was not about riders being dirty. It was about the arithmetic: cleaning happens between fares, every minute spent cleaning is a minute not earning, and the wipes that do the job attack the interior the driver owns and will eventually sell.
One of those drivers was Willie, and he became the person the whole ecosystem was designed against — not a composite persona assembled from a spreadsheet, but someone we could go back to when an argument inside the team needed settling. The film in section 08 follows his shift.
Willie was the reason the brief came out the way it did: make sanitisation something that happens while the car is working — including while someone is sitting in it. Any system that requires him to stop the car is a system that costs him the thing he came to work for, and he will stop using it in a week.
222 was built by Team Ultimanium. My work on it was:
- Industrial design of both modules — form, ergonomics, the cupholder envelope and the VHB ceiling mount
- RIDR app UX and UI across both the driver and rider flows
- The Far-UVC research and safety case, including the exposure-limit envelope the design had to sit inside
Three things that all had to be true at once.
Passive and autonomous
The cabin sanitises itself between rides. No wipe-down, no chemicals, no minutes lost between fares.
Safe while occupied
222nm is stopped by the outer layer of skin and the tear film, so the system can run with people in the car.
Visible to the rider
Sanitisation the passenger can't see is sanitisation they won't trust. The app surfaces it in real time.
The same event, told twice.
One sanitisation cycle is running below, and it does not stop when you switch sides. The driver gets a control surface and a number that climbs. The rider gets no number at all — because a percentage a passenger cannot verify is a claim, and the thing that actually reassures them is the reason behind it.
Watch the status bar at the top of the page
Operating the system between fares, and paid by the minute.
Modelled at 60× — 25 cabin-minutes in 25 real seconds. The curve is first-order inactivation, 1 − e^(−kt), with k fixed so a Deep Clean lands on 99.9% at 25 min. These are design targets. 222 was never deployed, so nothing here is a measurement.
Your Devices
Both modules · modelled 25 min to 99.9%
- ABUVActive: just now0% clean76%
Fixed to the headliner. Passive, and always the one already running.
Device info - NODEActive: idle0% clean52%
Lives in the cupholder. Directable, and oscillates through its arc.
Device info
The status bar at the top of this page is RIDR's, and it disappears on the rider side because it is absent from every rider screen in the shipped set. The gradient is the product's, darkened — as shipped, white on its cyan end measures 1.87:1, which is a bug in the app rather than a decision. Section 06 shows the measurements.
Both sides in full, as the live Figma prototypes — every screen, the real transitions, and the flow branching. You can walk either one yourself.
- The badge sits on the tier
- Far-UV appears at the moment of choosing a ride, not in a settings page. Trust has to be present at the decision.
- “Sanitised en route”
- The sentence the whole ecosystem exists to write. Dead time the driver used to absorb, shown as the rider's benefit.
- Citations, in a consumer app
- Further reading lists Columbia, Kobe and Nature Scientific Reports. Unusual — and right for a claim this counter-intuitive.
Two devices that aren't trying to be each other.
ABUV is fixed, overhead and passive. NODE is portable, directable and oscillates. A Deep Clean runs both together until the cabin reads 99.9% sanitised — which is the toggle you just watched flip on the console above.
Fixed, overhead, passive.
A ceiling-mounted far-UVC bar. The high vantage point is the whole argument — light reaches the footwells and seat backs that a device sitting on the console never sees.
Nobody presses anything.
The cabin sanitises across the whole shift, not in a slot carved out of it. ABUV runs passively at 222nm while the fare is in progress — which it is allowed to do because of the wavelength, not in spite of it. Deep Clean waits for the driver to step out, and it waits for a scheduling reason rather than a safety one: it takes 25 minutes, and nobody has 25 minutes between fares.
Phase 01, Fare in progress. 0 percent clean.
Targets, not measurements — 222 was never deployed. The 25-minute Deep Clean and the 11-minute fare are the project's own published figures; the four minutes between fares is an assumption, and the decay curve is the same first-order model the status bar at the top of this page is running.
The whole product rests on one number.
Conventional germicidal UVC at 254nm works, but it reaches living cells, so it can only run in an empty cabin — which defeats the point. Far-UVC at 222nm is absorbed by protein in the outer, dead layer of skin and by the tear film of the eye. That single difference is what lets the system run continuously, with passengers aboard.
- Stratum corneumdead cells
- Epidermis
- Basal layerliving, dividing
- Dermis
Absorbed by protein before it reaches a living cell. Safe to run while the cabin is occupied, within the exposure limit.
The published research is stated as dose, not time — a D₉₀ of roughly 1 mJ/cm² for aerosolised virus. Time-to-kill is a property of our fixture output and cabin volume, so this curve is our own model, not a lab result. The shipped app says the same thing in its own footnote: time to 99.9% varies with distance from the surfaces.
In 2022 the ACGIH raised the 222nm threshold limit from 23 to 161 mJ/cm² for eyes and 479 mJ/cm² for skin — the limit moved because the evidence strengthened, not because the claim was softened. Columbia's Center for Radiological Research found no skin cancer or abnormality after 66 weeks of chronic exposure, and a 2025 ocular study found no clinically significant photokeratitis.
Lethal to a virus. Nowhere near a person.
A cycle that kills 99.9% of airborne virus sits at roughly 2% of the exposure limit for human eyes. That ratio is the entire reason 222 can run with passengers in the car — and why conventional 254nm, which has no such margin, cannot.
Kill doses vary by pathogen and conditions; limits are the 2022 ACGIH threshold values. Both are dose figures — the time to reach them depends on fixture output and cabin geometry.
This page is wearing the product, and the product has four bugs.
Every colour on this page was sampled off the shipped RIDR screens rather than chosen. That is the only reason the next part is interesting: once real values go through a contrast gate, four of them fail — including the label on the status bar the driver reads in sunlight.
White on the cyan end: 1.87:1. Below every WCAG threshold there is.
Same hue, same direction, lightness moved until white clears 4.78:1.
| Token | Change | Why |
|---|---|---|
| signalFrom | #47CCF7 → #0F7BA4 | white on the shipped cyan scores 1.87; the status bar's own label is unreadable |
| signalTo | #1190F6 → #0B57C4 | white on the shipped blue scores 3.31; passes as large type, fails as a 13px label |
| accent | #0791FE → #0A6ECE | scores 3.24 on the white sheet; fine for the $8.50 figure, short for a link |
| nominalOnSheet | #34C759 → #12793C | “Active: just now” is 12px green on the white device sheet, and scores 2.22 |
These are findings about RIDR, not decisions about the case study. The page darkens four values so it can be read, prints the deviations at the bottom of every audit run so they stay visible rather than becoming folklore, and leaves the originals in the swatches above. The fix belongs in the app.
A day in the life of Willie.
The driver from section 01, on a shift with 222 installed. The film shows the ecosystem being used rather than explaining it — which is the only honest test of a product whose entire claim is that it disappears into the work.
Research, teardowns, sketches and the safety case — in full.
The complete document behind the ten weeks of IACT 330, including the interviews with Willie and the dose modelling.
222 was a six-person project. Naming everyone is the only honest way to present team work, and it costs nothing to be accurate about which parts were mine.
- Quint Bailey
- Charlie Bowles
- Guy Cassieri
- Sathvik Kotha
- Wyatt Michel
- Nico Zafarana
222 was a concept. Farm OS is the same instinct, shipped.
Sensors in an enclosure, telemetry on a phone, and a system that has to keep working when nobody is watching it — running on a real farm.
Farm OS