Safety Infrastructure / Founder Concept
Kaleido
A paired roadside system designed to keep temporary work-zone traffic in a verified, supervised state.
Temporary traffic control still places people beside moving vehicles, often with limited visibility and two ends of a one-lane closure that cannot see each other. Kaleido began with that physical, human problem - not with a request to add AI to a sign.
The resulting concept is a paired, field-deployable roadside system. Each unit observes its local approach, a deterministic safety protocol checks clearance and device health, and a device-to-device handshake allows one direction to proceed only when both ends agree. Any uncertainty returns both units to STOP.
I led the field discovery, product strategy, safety architecture, industrial design, prototype direction, patent work, and operator experience. This is a feasibility-stage prototype and research program, not a certified traffic-control device or evidence of public-road deployment.

01 / field discovery
The issue was never simply a better STOP sign.
Early conversations with roadway, utility, and temporary traffic-control workers pointed to a coordination problem. Teams need to keep traffic predictable while managing short staffing, glare, changing geometry, and a work zone whose two ends may not share a line of sight.
That reframed the brief. Kaleido should help a crew supervise traffic from a safer position, retain human oversight, and make the shared condition of the zone legible. It should not claim to replace trained flaggers.
More than 10 early field interviews and roughly 20 feedback points became product requirements - not decorative research quotes.
Exposure is built into the role
Move continuous supervision away from the edge of live traffic while keeping an operator in the loop.
Two ends need one shared truth
Treat synchronized local state as the core product behavior, rather than ask one device to decide alone.
Signals can be misread in the field
Make release conditional on verified clearance and an explicit paired handshake - not on a display alone.
Field equipment must earn its footprint
Design the physical system around portability, setup, visibility, stabilization, and operator access.

The product goal is not to replace a flagger. It is to reduce unnecessary exposure while preserving accountable supervision.
02 / design brief
I translated the field problem into three product rules.
The category already includes portable signals and automated flagger assistance devices. Kaleido does not need an unfamiliar visual language to be useful. Its contribution is a protocol that makes the two ends of a bounded, supervised work zone agree before traffic moves.
Those requirements kept the industrial design honest: the system has to be legible to drivers, serviceable by a crew, stable at the roadside, and compact enough to travel as part of temporary work-zone equipment.

No single point of release
One camera, one unit, or one ambiguous perception result must never independently release traffic.
Rejected: a smart sign that acts on a single local observation.Safety logic stays local
The release path is designed around on-device checks and a paired handshake. Cloud tools remain for monitoring and records.
Rejected: a cloud-dependent safety path.The worker owns the system
Setup, reset, manual hold, and emergency override remain explicit physical and interaction responsibilities.
Rejected: an autonomous-worker narrative.03 / paired architecture
The technology is a protocol, not an AI camera.
Each Kaleido unit performs local optical sensing and produces a constrained clearance classification. The product separates perception from authority: edge AI can suggest CLEAR, OCCUPIED, UNCERTAIN, or FAULT, but it cannot release traffic by itself.
A low-power, device-to-device LoRa handshake carries state, confidence, direction, heartbeat, and safety status between the two endpoints. The cloud can support monitoring, event logs, and reporting, but safety-critical decisions are held locally.
- 01Sense locally
- 02Verify clearance
- 03Handshake
- 04Release one side
- 05Log and reset

Local logic holds the line. Two endpoints verify. An operator owns setup, reset, and emergency override.
04 / safety states
The most important interaction is knowing when not to proceed.
STOP is both the startup state and the fallback state. Before a direction can be released, both units must pass health checks, classify the zone as CLEAR, exchange compatible state, and agree on direction. One unit shows SLOW or CLEAR while the other remains STOP.
Communication loss, camera obstruction, low confidence, a person or object in the zone, device tilt, critical battery, state mismatch, or a manual override all return both units to STOP. That is a design decision with physical, software, and operator consequences.

Default ALL STOP
Power-up, idle, completion, and any exception begin from a state that cannot release traffic.
Paired state handshake
Compatible local CLEAR results must be synchronized before the selected direction can move.
Fault STOP
Uncertainty is a product state, not a hidden error. It is logged and requires review or reset.
For a safety-critical product, the right failure behavior is part of the core interaction design.
05 / form to field
The form had to integrate seven jobs without looking like a robot costume.
The product combines a double-sided display, elevated sensing, solar-assisted power, an adjustable mast, a sealed control enclosure, stabilization legs, and transport wheels. I treated these as a physical sequence rather than an inventory of features: arrive, unfold, stabilize, align, pair, supervise, and pack down.
Early physical studies made the trade-offs visible. The display needs road-facing authority; the camera needs a clear view; controls need protected access; and the base must remain stable while giving one person a credible transport path.



In safety hardware, form factor is part of technical feasibility.
06 / component architecture
Every visible component carries a field requirement.
The final arrangement separates public-facing communication from protected service work. The sign head and side lights remain readable at a distance; the camera and solar plane sit above the traffic message; status and power controls live at the base; and the tripod opens below the unit's center of mass.
This architecture lets the product present a simple roadside message while exposing a clear maintenance and setup logic to the crew. It is a more disciplined alternative to hiding a complex system inside a generic black box.



The product should be instantly legible to traffic, and deliberately legible to the person who has to set it up.
07 / final industrial design
The final direction makes the system feel like equipment, not speculative machinery.
The production-direction render resolves the early questions into a restrained, rugged product language: a black enclosure for roadside contrast, an orange service and articulation layer, a clear vertical stack, and a base that signals both stability and mobility.
This is not a cosmetic final-render chapter. It is the point at which the design's roadside message, maintenance logic, power geometry, and transport posture become one coherent object.



The final form comes from the workflow, not the other way around.
08 / setup and mobility
Deployment begins at the truck, not at the sign face.
For a temporary work-zone product, the off-road condition is as consequential as the deployed condition. The display folds into a protected transport configuration; the mast, solar plane, and legs pack down; and the wheels plus pull handle make one-person movement an intentional part of the product architecture.
The sequence studies are not beauty shots. They identify the exact transition that needs further validation: how a crew lifts, opens, levels, aligns, pairs, and packs down the unit under real field constraints.
- 01Roll to position
- 02Unfold
- 03Stabilize
- 04Raise
- 05Pair
- 06Supervise
- 07Pack down



Portability is not a convenience feature. It is a safety and adoption requirement.
09 / supervised operation
The interface is a set of visible safety states - on the hardware first.
The roadside unit communicates power, synchronization, zone-clear status, and traffic instruction without requiring an app. The companion interface supports setup, health checks, manual override, logs, and supervised monitoring, but the essential operating state remains readable on the physical equipment.
The visual language has redundancy by design. The central message board, side lights, and audible cues help approaching drivers and nearby crews recognize a stop condition or fault without treating a mobile screen as the source of truth.
- 01Deploy
- 02Pair
- 03Align
- 04Verify
- 05Supervise
- 06Manual hold
- 07Power down



A safety device should never make its current state a guessing exercise.
10 / validation plan
The next milestone is evidence, not autonomy.
The validation plan starts in a controlled, low-speed, single-lane environment. It measures detection performance, false-clear events, communication timeouts, STOP fallback, display visibility, setup time, alignment errors, battery behavior, and operator recovery from faults.
Success is not a polished demo. It is a documented safety envelope, known failure modes, reproducible state transitions, and enough field evidence to decide whether a supervised pilot is responsible.

Instrument the normal path
Measure paired handshake latency, state transition correctness, display legibility, and setup time in a controlled setting.
Inject realistic faults
Test communication timeout, camera obstruction, low confidence, intrusion, battery degradation, tilt, and state mismatch.
Make STOP the measurable outcome
Evaluate whether uncertainty reliably returns both units to a documented, reviewable STOP condition.
For Kaleido, the most important feature is the condition under which it refuses to proceed.