HQZN / EDGE INTELLIGENCE
Edge intelligence
Bringing local intelligence closer to equipment and on-site work.
We research edge computing and on-device inference for field equipment, exploring how data processing supports intelligent applications while validating reliability and deployment requirements.
Explore our R&DPRODUCT DEVELOPMENT
Built around the work.
Equipment developers, charging operators and intelligent product teams
- Device data collection and protocol integration
- On-device inference and application validation
- Edge computing and deployment research
SYSTEM WORKFLOW
Workflow
Equipment
Edge processing
Inference
Business APIs
Development status
Embedded-board and site-level inference experiments have been completed. Perception, interaction and deployment configurations are validated separately, not claimed as fleet-wide production capabilities.
One task. Different edge configurations.
Comparing local open-model inference across platforms and quantization configurations to inform on-device application development.
- Development platforms
- NVIDIA Jetson AGX Thor; NVIDIA DGX Spark for development and cross-validation.
- Tasks
- Code, mathematics and text generation, with fixed model and software versions, repeated runs and output checks.
- Method
- The common protocol includes 3 warm-up runs and 5 measured runs per test, with numerical probes and fixed-corpus quality evaluation where supported.
- Output
- Platform- and configuration-level compatibility and performance records to inform workload-specific inference choices.
Validation scopeLimited to the tested versions and single-request experiments. Some configurations lack full compatibility or quality validation. Concurrency, long-context and end-to-end vision were not validated in this study; it does not establish fleet-wide production readiness.
These third-party names are trademarks or registered trademarks of NVIDIA Corporation and identify platforms used in our research only.
PILOT SCOPE
Discuss a pilot
Inputs to align
Device resources and operating environment, task samples, response requirements and data boundaries.
Potential project deliverables
- Local inference prototype
- Configuration and compatibility records
- Evaluation summary for an agreed task
Scope is agreed after reviewing data, equipment and acceptance criteria. Research validation, pilots and production delivery are defined separately.
LET’S BUILD WHAT’S NEXT
Let's talk equipment, data and operations.
Discuss device integration, operations software, on-device inference or digital twins with our team.