A dancer with a prosthetic leg, arched back on a reflective floor
Viva Technology

Physical AI

Skill never dies. It compounds.

Evelyn Mora

We Are French Touch – Paris

Tech can’t scale without
Fashion, Art and Entertainment.

A VLGE retail world with its live analytics and visitor heat map
Space
Models standing on a runway under spotlights
Experience
A grey tailored jacket dress with sculpted shoulders
Design
Two guests at a party in sunglasses
Culture

We further fine-tune our data with select consented expert data.

VLGE

We build and capture the world.
Then the skill within it.

The world’s most guarded rooms are our training data.

The Empire State Building
The palace and gardens of Versailles
The Louvre pyramid at dusk
The Hall of Mirrors, Versailles
An observation deck under a mirrored ceiling
The Metropolitan Museum of Art
SpacesTens of thousands of 3D, Gaussian splat and 360° worldsMuseums · luxury houses · fashion weeks · retail
3M hoursof build data: how worlds are made, by amateurs and pros
4M+ hoursof play data: how people move, decide and transact
Expert skillConsented, instrumented human talent in real spaces

VLGE · Spatial and human intelligence for physical AI

Today · The data at work

Retail runs on how people move.

A VLGE store world: traffic paths and heat zones on the floor, analytics panels around them
  • MerchandisingWhat gets picked up, compared and bought, through to the transaction
  • RetailTraffic flow, dwell time, conversion by zone
  • Interior designLayouts tested in the world before they are built
  • Floor plansHot zones, dead zones, the paths people actually take

Brands fund the worlds and get the analytics · VLGE licenses the anonymized data

Physical AI

Robotics is a systems problem.

Today

Factories and warehouses

  • Robots rented by the hour
  • Optimized for throughput and unit cost
  • The bottom line decides
Next

Art, luxury, entertainment

  • Judged on execution and taste
  • Every output carries a brand
  • Experts define what good means

Harvey put lawyers in the loop of legal AI.
Physical AI needs the same for craft.

VLGE · We Are French Touch

The science of taste

Taste makes something ugly, silly, beautiful.

Perceptiontexture, form, color Emotionreward, risk, desire Meaningculture, memory, story Taste
Appeal Novelty, risk → Modelsthe safe average Tastethe edge that still lands Noise

Illustrative. After the aesthetic triad (Chatterjee & Vartanian, 2014) and the arousal curve (Berlyne, 1971).

The future of data

Tomorrow, value lives in branded data.

Today

Spaces

Museums, stores, homes

Today

Behavior

How people move, decide, buy

Next

Expertise

The skill of a dancer, an artist, a craftsman

Frontier

Branded data

The taste and essence of one named expert. Consented, attributed, licensed like IP.

Generic data trains capability · branded data trains taste

Expert capture · Ballet

Timing, balance, expression.

A dancer in a deep arch, one hand and one pointe on the floor Wrist IMU Gaze · spotting Joint angles Floor contact Pointe pressure
  • BodyFull-body joint angles, balance, weight transfer
  • ContactFoot, pointe and hand pressure on the floor
  • TimingEvery movement locked to the music
  • Gaze and effortSpotting in turns, EMG, breath

Consented demonstration → structured multimodal data

Expert capture · Makeup

Pressure, angle, color judgment.

VLGE expert capture · retail luxury, makeup store

PressureBrush and fingertip force on skin
AngleBrush tilt, rotation, approach to the face
StrokeSpeed, direction, blending passes
GazeWhere the eye checks: face, light, palette
Color judgmentShades picked, rejected, mixed, and why
ContextSkin tone, lighting, the brief, the story

Consented demonstration → structured multimodal data

Brain and body data

The judgment behind the hand.

A woman wearing an EEG cap
EEG · decision states
EMG · muscle effort
Heart rate · EDA
Gaze · attention
Hands · fingers

Shared capture schema · vision, kinematics, physiology, context · 415 fields

Expert to agent

Instrument the person, the task, the space.

01
Experts in their domain
02
Skill-specific signals
03
Skill transferred to the agent

Pink human reference · orange agent

Agents at scale

One expert. A thousand runs.

run 1
run 2
run 3
run 4
run 5
run 6
×N
The agent re-runs the task ×NScaled data that stays human-like
The agent generates the long tailRare scenarios, measured against the human

VLGE human-grounded agents · trained on real human behavior

From capture to model

One capture. Three ways to train.

Dataset
Fine-tune agents
Train robot policies
Benchmark
Ballet
Motion agents that generate expressive movement for film, games and avatars
Humanoid whole-body control: balance, recovery, weight transfer
Behavioral-realism score: does it move like a dancer, or only reach the pose?
Makeup
Beauty agents that reason about undertone, light and face shape
Applicator robots with force and angle control safe near the face
Expert-ranked outputs: the artist scores the look, the model learns why
Any craft
Agents that carry a named expert’s method and voice
Dexterous hands trained on real touch, not teleop guesses
The expert as ground truth for every lab’s model

Capture once. License, train and benchmark many times.

VLGE Human-Grounded Benchmark · real human capture as the reference

We Make Talent Profitable1/2

Expert talent

World-class human expertise
ChefsCulinary techniques
AthletesMovement & performance
ArtistsCreative process
DesignersSpatial & product design
PerformersMusic, dance & expression
SpecialistsDomain-specific skills
Real experts. Real skills.
Real-world scenarios.
Portrait of an expert in a blue shirt
Portrait of a chef in whites
Portrait of an artist
Portrait of a designer in tortoiseshell glasses
Portrait of an athlete in a blazer
Portrait of a performer

Immortal and
Licensable
in Perpetuity

Prime agents

From Experts, To Prime Agents.

A chef cracking an egg into a bowl, with hand landmarks traced
The chef's eyes, close up
The chef's hands, close up
Eye gaze &
attention
Hands &
fingers
Audio
Brain
activity
Transfer Skills From
Expert to Agents
A humanoid robot taking a soufflé out of an oven
Soufflé
Scrambled eggs on toast
Eggs Benedict
Omelette
Omelette with salad and a coffee

Shared Capture Schema(applied to all domains)

Xd(t) = [Wd(t), Bd(t), Hd(t), Ad(t),
Zd(t), Sd(t), …]
∀ d ∈ 𝒟,  Xd(t) ∈ Sshared
  • W= world & object state
  • B= body pose
  • H= hand kinematics
  • A= action stream
  • Z= gaze & attention
  • S= spatial sensors
  • …= additional streams
  • 𝒟= set of domains
  • Sshared= shared schema
    space (415 fields)

Shared extracted features(applied to all domains)

φshared = Ψshared(Xd)
  • trajectory smoothness
  • speed profile
  • dwell time
  • action frequency
  • ⋮

Skill-specific derived signals(domain-specific)

φdskill = Ψd(Xd)
Chef (knife skills)
  • blade speed
  • cut angle
  • grasp stability
  • eye-before-cut
  • …
Sport (climbing)
  • CoM rate
  • route-read latency
  • grip force
  • slip risk
  • …
Final representation φd = [φshared; φdskill] Different domains derive
different skill signals
φchefskill ≠ φsportskill

With Human Prime Ground Truth Benchmark
to capture Branded IP and Copyright2/2

Proprietary human data

Expert skill captured as multimodal data
Body MotionFull-body kinematics
HandsFinger motion + manipulation
ContactTouch + force
GazeAttention & visual intent
Cognitive SignalsDecision states + errors + retries
EnvironmentScene + object context
Multimodal. High-fidelity.
Real-world context.

Expert IP

Human expertise becomes
structured, reusable intelligence

CaptureMulti-modal
fusion
StructureBehavior
representation
LearnSkill
modeling
From human performance
to machine capability.

Licensable intelligence

One performance. Reusable in perpetuity.
LicenseProprietary training data
for AI & robotics
BenchmarkMeasure AI vs. expert
human behavior
TrainTask-specific agents
on expert skills
ScaleExpert skills across
machines and environments
Immortal skills.
Real-world impact.
The economy of expertise

Skill never dies.
It compounds.

Capture License Train Benchmark Paid in perpetuity
Experts stay authorsTeachers and evaluators of every model trained on their craft
Likeness is an assetVoice, motion and essence: consented, attributed, compensated
Savoir-faire, machine-readableFrance holds the craft. Now it can license it.

Evelyn Mora · evelyn@vlge.com · vlge.com