Core Technologies
Cross-Modal Data Fusion
Cross-modal feature alignment and semantic fusion across heterogeneous data (temporal, spatial, semantic dimensions)
Bio-Contextual Modeling
Coupled modeling of mind–body states (emotion, stress, fatigue) and contextual attributes (location, activity, social environment)
Advanced Neural Architectures
Attention-based architectures, graph neural networks, and temporal sequence modeling
Evolutionary Trend Prediction
Long-horizon behavioral modeling and evolutionary trend prediction
Dynamic Sensor Integration
Dynamic plug-in multimodal integration mechanism for heterogeneous sensor access
Technical Innovation
One of the first implementations of high-order multimodal fusion directly on wearable-edge systems, enabling robust interpretation under complex real-world conditions
Contactless Continuous Human State Perception
Millimeter-Wave Radar Physiological Sensing Platform
ThingX pioneers the integration of millimeter-wave radar into next-generation wearable sensing systems.
Core Technologies
- Non-contact extraction of heart rate, respiration, and micro-movement signals
- Clutter suppression and micro-motion signal separation algorithms
- Fusion of low-information-density sensors (IMU) with high-density audio sensors
- Modular, low-power, miniaturized sensing architecture for long-duration deployment
- Exploration of next-generation sensor expansion for broader wearable scenarios
Technical Innovation
Establishing a privacy-preserving, contactless physiological sensing paradigm for real-world continuous monitoring.
Near-Sensor Edge AI Computing Architecture
ThingX develops a near-sensor computing architecture that minimizes cloud dependency while maximizing energy efficiency.
Core Technologies
- Heterogeneous computing architecture (CPU + NPU + DSP)
- Local perception–preprocessing–inference closed-loop pipeline
- Dynamic task scheduling and power-aware computation
- Edge–cloud collaborative load balancing
Technical Innovation
Transforming the traditional “sense → transmit → cloud compute” paradigm into
“sense and compute locally.”
This architecture enables scalable deployment of AI intelligence on resource-constrained devices.
Core Technologies
Model Compression
Structured pruning, low-rank decomposition, knowledge distillation
Adaptive Inference
Mixed-precision quantization and adaptive inference
Architecture Synergy
Hardware–model co-design (NAS + hardware optimization)
Edge AI Engine
Dedicated multimodal edge inference engine
Technical Innovation
Enabling efficient execution of hybrid and large models on edge devices, paving the path toward wearable-scale AI deployment.
Continuous Intelligence Without Raw Data Exposure
Distributed Model Evolution & Privacy-Preserving Learning
ThingX develops distributed model iteration frameworks that preserve privacy while enabling continuous system evolution.
Core Technologies
- Federated learning and incremental adaptation
- Communication-efficient distributed training mechanisms
- Edge-collaborative parameter updates
- Differential privacy and secure multi-party aggregation
Technical Innovation
A device–edge–cloud collaborative learning architecture capable of large-scale multimodal evolution while protecting user data sovereignty.
Long-Context Proactive Interaction Architecture
ThingX explores proactive interaction frameworks driven by long-context perception and intent inference.
Core Technologies
- Cross-session long-context modeling
- Dynamic interaction strategy generation
- Multichannel feedback (voice, light, haptics)
- Integration with smart homes, smart mobility, and ambient computing systems
Technical Innovation
Redefining the role of intelligent systems —
from passive responders to context-aware adaptive companions.