Scalp-safe approach
Establish a live scalp-relative datum, maintain a defined force ceiling, tolerate human movement, and prevent unsafe motion independently of hair capture.
RAE / Robotic Aesthetic Engineer
Robotic haircutting is a soft-body manipulation problem.
RAE is an autonomous haircutting system being developed to safely perceive, capture, stabilize, and cut human hair on a moving person.
Designed from more than three decades behind the chair. Now being translated into sensing, controls, mechanics, and machine intelligence.
Current stage: architecture complete · bench validation underway · technical co-builder needed
01The real problem
Hair does not behave like a manufactured part.
It bends, stretches, tangles, compresses, changes with moisture, and moves differently across density, texture, curl pattern, length, and prior damage.
The head beneath it also moves.
A robotic haircutting system must do more than recognize a hairstyle. It must continuously understand where the person is, determine what the machine is touching, control hair under tension, and prevent cutting motion whenever reality disagrees with the model.
- 01Soft-body manipulation
- 02Computer vision
- 03Contact + force sensing
- 04Compliant robotics
- 05Real-time controls
- 06Functional safety
- 07Computational haircut design
The haircut is the visible result. Reliable physical interaction is the actual invention.
02System architecture
RAE separates perception, positioning, hair control, and cutting.
Each layer earns the next layer’s authority.
- 01
Plan
Translate haircutting knowledge into structured geometry: shape, length, elevation, direction, boundaries, and sequence.
- 02
Perceive
Use RGB, depth, and spatial sensing to estimate the head, hair, tools, environment, and human movement.
- 03
Establish live truth
Confirm the current physical relationship between machine and person through compliant contact and force sensing.
- 04
Capture and stabilize
Handle selected hair as a controlled bundle instead of attempting to cut loose hair in open space.
- 05
Permission
Require independent agreement across live contact, force, proximity, containment, tool state, position, and human veto channels before granting temporary, revocable cutting authority.
- 06
Contained cut
Operate only inside a mechanically bounded and verified path after permission has been granted.
- 07
Verify and log
Monitor force, motion, proximity, sensor agreement, interruption states, and execution results.
The map is a belief. Live touch is truth. Steel never moves on the map alone.
Vision guides the system. It does not receive sole authority to move a cutting tool near a person.
03AetherFlow
Control the hair before attempting to cut it.
AetherFlow is RAE’s hair-handling architecture. Its purpose is to gather selected hair, separate it from uncontrolled material, establish direction and tension, present it at a defined cutting plane, contain the cutter’s path, and evacuate severed fibers.
The current work is not about maximizing suction. It is about producing a repeatable mechanical state from material that does not naturally want to be repeatable.
01 / 06
Selected field detectedAcquire the material without implying control.
CUT AUTHORITY / WITHHELDBEHAVIORAL MODEL Non-dimensional public view / implementation protected
- 01Gather
- 02Isolate
- 03Align
- 04Stabilize
- 05Cut
- 06Evacuate
RAE does not ask a robot arm to recreate a human hand. It redesigns the task so the machine can perform it safely.
04Current work
We are proving the dangerous assumptions before building the beautiful object.
RAE is in the bench-validation stage. There is no functioning autonomous pod today. The finished pod is the product vision. The present work is the machinery and evidence required to earn it.
Evidence register / public view
System architectureDocumentedPublic doctrine
Permission logicStress-tested in simulationControlled review
Physical explorationActiveControlled review
Minimum observable rigDevelopingConfidential implementation
Deliberate boundary
The work is visible. The replication path is not.
Mechanism geometry, dimensions, tuning, sensor thresholds, control rules, and raw validation material are shared only with vetted technical collaborators.
Join the build ↗What exists
- Complete product + system architecture
- Defined sensing, control, mechanical + safety layers
- AetherFlow capture + contained-cut architecture
- System specifications + interface definitions
- Failure-mode + safety analysis
- Gather–lift–cut control simulator
- Hostile-condition testing of permission logic
- Physical airflow + mechanical exploration
- Developing minimum observable validation rig
- 35+ years of haircutting knowledge entering machine logic
Current proof sequence
Hair control
Demonstrate repeatable capture, isolation, alignment, and stabilization of a defined hair bundle.
Contained cut
Demonstrate that stabilized hair can be severed accurately inside a mechanically bounded operating path.
Integrated repeatability
Combine perception, contact, handling, cutting, evacuation, interruption, and logging into a measurable bench cycle.
We are not building the pod first. We are building the reasons the pod deserves to exist.
05Engineering challenges
The problems currently on the table
Scalp-relative localization
Maintain an accurate relationship to a deformable, partially obscured surface while the person moves.
Hair versus scalp contact
Determine whether the end effector is interacting with free hair, compressed hair, scalp, ear, clothing, or another object.
Compliant contact
Maintain useful contact without rigid positioning or exceeding safe forces.
Soft-material capture
Convert inconsistent loose fibers into a stable, measurable bundle without damage or over-tension.
Bounded cutting
Prevent the cutting mechanism from operating outside its contained and verified path.
Sensor disagreement
Define behavior when vision, depth, force, position, and contact signals tell different stories.
Hair diversity
Create repeatable behavior across texture, density, moisture, length, and prior damage.
Verification
Turn “it appears to work” into instrumented, repeatable, falsifiable evidence.
06Safety / permission logic
Anything uncertain means no movement.
RAE begins locked. Physical containment, live sensing, tool readiness, and human veto must independently agree before one temporary cutting action can exist. A person can revoke that authority at any moment.
01 / 07
Locked by default
Nothing moves until the system earns authority.
07Who we need
RAE has reached the point where the next breakthroughs must happen in hardware.
The current gap is not another concept designer, branding agency, or person who believes robots are neat. RAE needs technical people who want to help prove the hard parts.
Primary need
Technical co-builder
A long-term engineering partner with meaningful experience in some combination of:
Embedded systemsSensors + instrumentationControlsMechatronicsRoboticsCompliant mechanismsForce-controlled interactionMachine visionSafety-critical hardware
The right person does not need to know hair. They need to be curious about unfamiliar physical problems, comfortable building crude evidence before polished products, and willing to challenge assumptions without killing the ambition.
Someone who likes an ugly working rig more than a gorgeous render.
Also interested in hearing from
- Robotics + computer-vision researchers
- Mechanical + mechatronics engineers
- Sensor + actuator manufacturers
- Safety + human–robot interaction specialists
- University labs + student engineering teams
- Technical advisors for early physical systems
08Founder
Since 1989
Built from the chair outward.
RAE was founded by Thad Grace, a hairdresser, artist, builder, and Brazilian jiu-jitsu black belt based in Oregon.
Thad started hair school in 1989 at 17 years old. Decades behind the chair taught him that haircutting is not primarily about the blade. It is the constant work of reading material, controlling tension, responding to movement, protecting the person, and making hundreds of small decisions under uncertainty.
RAE began with a question: could that embodied knowledge be translated into a machine without stripping away the care that makes the work human?
“The goal is not to make a robot imitate a hairdresser. The goal is to understand the work deeply enough to redesign it for a machine.”
Build the evidence
Help prove that autonomous haircutting can be safe, repeatable, and real.
RAE is not currently selling pods. It is building the technical foundation required to make them possible.
We are looking for engineers, researchers, collaborators, and one serious technical co-builder who sees the size of the problem and wants to work on it anyway.