Press Cmd/Ctrl + P to export: choose Landscape, set Margins to None, and enable Background graphics to preserve the full visual design.

Back to Deck
Abstract physical world represented by spatial particles and trajectories
Investor deck · 2026

PhysenseAI

Physical World Understanding Model

Turning physical observations into machine-understandable reality.

Physical intelligence begins with understanding
02The Next Model Frontier

AI can understand digital information. Now it needs to understand the physical world.

Language
“The cat is on the roof.”
LLM
Images / Video
Pixels · Frames
Vision Models
Physical World
Signals · Time · Change
?
Category revealPHYSENSEAI

This is not another sensing device. It is a new model layer.

03Observation ≠ Understanding

Observation is not understanding.

Fragmented observations
?

What actually happened?

OBSERVEEVENTSTATEMEANING
Machine-understandable event

When AI receives only fragments, software and machines make decisions from an incomplete reality.

04Why Now

Four curves are converging on one missing layer.

01 · RICHER SENSING

Physical signals are richer and continuous

02 · EDGE COMPUTE

Continuous inference can run locally

03 · AI LEARNING

Temporal and multimodal learning is maturing

04 · PHYSICAL AI

AI is moving from screens into reality

UNDERSTANDING
the missing layer

The world can now be observed at scale. The missing layer is understanding.

05Perceptual Map

A new position in the AI stack

Signal detection → Physical understandingTask-specific → Generalisable
Sensor vendors
IoT platforms
Camera AI
Robotics perception
PHYSENSEAI
The position

From detecting a signal to forming a reusable understanding of reality.

This defines a model category—not a competitor list.

PhysenseAI does not optimise one task. It builds a physical representation reusable across tasks.

06The Physense Model

From heterogeneous observations to one physical representation.

Signals converging into a unified physical representation
Motion
Pressure
Vibration
Audio
Vision
Environment
Unified · Structured · Learnable
PHYSICAL
REPRESENTATION
Events
States
Context
Dynamics
Prediction

Not an IoT pipeline. Not a software stack. The core asset is a learnable, reusable Physical Representation.

07Semantics → Representation Learning

We are in the first stage of AI model development.

TODAY

Interpretable Physical Event Inference

01Real physical observations
02Event ontology
03Structured supervision signals
04Validated physical semantics
NEXT

Learned Physical Representation

01Supervision + real-world sequences
02Learned event / state representation
03Cross-environment generalisation
04Prediction

Rules are not the destination. They are a path to physical semantics and structured supervision.

08What PhysenseAI Learns

Understanding is not a label. It is a growing capability.

01
EVENT
What happened?
02
STATE
What is happening?
03
CONTEXT
What does it mean?
04
DYNAMICS
How is it changing?
05
PREDICTION
What happens next?

One learned representation can serve software, intelligent spaces and Physical AI.

09Evidence & Validation

A clear boundary: proven, next, and research.

PROVEN01
Technical validation
  • Real-device integration
  • Multi-source time alignment
  • Event reconstruction
  • Continuous / edge inference
Initial continuous real-device validation dataset: 41 hours
NEXT02
Real-world validation
  • Longer continuous sequences
  • Cross-environment stability
  • Error analysis where ground truth exists
RESEARCH03
Learned generalisation
  • Learned representation
  • Beyond interpretable baselines
  • Cross-environment generalisation

Technical validation is underway. Commercial validation remains early market discovery.

10The Research Thesis

Turn invisible reality into a learnable loop.

A circular physical world learning loop
01 · REPRESENT
Can we represent physical reality?
02 · RECONSTRUCT
Can we reconstruct from incomplete observations?
03 · GENERALISE
Can learning move beyond rules?
PHYSICAL WORLD
Observe · Represent · Reconstruct · Learn · Generalise

Every research question must resolve into validation, a deliverable and a new capability.

11One Representation · Many Environments

From one real-world wedge to many environments.

ONE PHYSICAL
REPRESENTATION
Eldercare & Care Centres
First validation and commercial wedge
BEACHHEAD · FIRST PAYING MARKET
Intelligent Spaces
Buildings and places that need live context
Industrial Environments
Complex processes, machines and people
Physical AI
Machines that must understand before acting

Prove the model and business loop in care environments first—then reuse the same physical understanding across environments.

12Market Opportunity

From care-institution beds to a serviceable software market.

Start with eldercare and care-centre operators in China. Other environments remain expansion vectors—not double-counted.

TAM
RMB 6.1B / YR
5.077M registered care-institution beds
× RMB 1,200 / bed / year
SAM
RMB 4.0B / YR
65.7% are nursing-care beds
≈3.336M nursing-care beds
SOM · YEAR 5
RMB 20M ARR
0.5% of SAM beds
≈16,700 paying beds

Source: Ministry of Civil Affairs / China National Committee on Ageing, 2024 Statistical Bulletin: 40K registered care institutions, 5.077M beds; 65.7% nursing-care beds.

Founder assumptions: RMB 1,200 annual software value per bed; year-five SOM at 0.5% of nursing-care beds. Deployment / integration fees excluded.

This is a bottom-up wedge thesis—not the company’s terminal market or a revenue forecast. Paid pilots will calibrate pricing and penetration.

13Why This Becomes Defensible

Every exposure to the real world makes the model harder to copy.

Compounding physical world learning loop
Compounding core
PHYSENSEAI
GETS BETTER WITH
REAL-WORLD EXPOSURE
More real environments
More physical sequences
Better physical representation
Broader generalisation

The moat is not a static dataset. It is a deployment–learning–validation system that compounds.

14Commercialisation & Initial Wedge

Solve a clear care-operator buying trigger—then expand the model value.

First paying market
Eldercare & care-centre operators

Buying trigger: understand room events and states continuously, reduce care blind spots, and support proactive checks and exception response.

First buyers
Care operators · Facility directors / Operations leads
Channel
Direct sales · Care-operation partners · System integrators
DEPLOY / INTEGRATE
Connect rooms and existing sensing in care institutions
ANNUAL SUBSCRIPTION
Per-room or per-site event, state and context understanding
API / SDK
Open to third-party software and more environments after validation
Current stageEarly customer interviews · Paid-pilot thesis to validate

Customers pay for continuous, explainable room-state understanding—not for more sensors.

15Team + Funding Ask

18 months to build the core team and prove a learnable Physical World Understanding Model.

Yang Li · Founder

MSc in Information Processing Science; 17 years in technology with repeated 0→1 product execution. AI-native product, agent and SaaS experience; responsible for translating R&D into international products and commercial pathways.

Henglin Shi · Scientific Advisor

Assistant Professor in Computer Vision and Learning Systems at Linköping University; PhD from Oulu and postdoctoral research at KTH. Focused on human behaviour understanding and multimodal signal processing.

NEXT CRITICAL HIRE
CTO / Head of AI

Full-time ownership of model R&D, physical representation learning, training and validation—and building the core AI engineering team.

What are we raising?
USD 1.4M
Angel / Seed
0–6 MONTHS
Technical lead hired · representation baseline
6–12 MONTHS
Real-world reconstruction · data & error analysis
12–18 MONTHS
Learned generalisation on real-world sequences
CAPITAL
40% model & representation · 35% real-world validation · 25% team & operations
PROOF
Physical representation · Real-world reconstruction · Learned generalisation
UNLOCK
A learnable Physical World Understanding Model

We know the team gap—and this round funds the full-time model leader needed to close it.

Abstract physical space formed by particles and trajectories
PhysenseAI · Physical World Understanding Model

Give AI an understanding of what is happening in the physical world.

PhysenseAI+86 136 1190 9946contact@physense.world