Wenqi Wang / Physical-AI engineer · FounderWenqi Wang / Physical AI 工程师 · 创业者
I explore AI-native ways of learning and building, with a focus on how Physical AI connects sensing, compute and action. From agents to embedded systems, I test ideas on real hardware and log every experiment. 我探索 AI-Native 的学习与创造方式，关注 Physical AI 如何连接感知、计算与行动。从智能体到嵌入式系统，在真实硬件上验证想法，记录实验与成长。
Sydney, AU · 33.87°S 151.21°E · building Aurai Sensing澳大利亚悉尼 · 33.87°S 151.21°E · 正在打造 Aurai Sensing
Digital self · 3D scan v5
02 / Profile
I’m an instrument engineer turned Physical-AI builder. I trained in Measurement & Control Technology and Instruments at China Jiliang University, the discipline of turning the physical world into data you can trust. I’ve been designing and soldering my own boards since my first STM32 minimum system in 2021.
我是一名从测控仪器走向 Physical AI 的工程师。本科就读于中国计量大学测控技术与仪器专业（光电检测方向），这门学科的核心是把物理世界变成可信的数据。从 2021 年自己设计并焊接第一块 STM32 最小系统板开始，我一直在真实硬件上动手。
Since then I’ve built picking robots, smart-car vision, FPGA-accelerated CNNs, a pipeline leak-tracing robot and Sub-1GHz sensor networks for agriculture, and documented all of it in my Lab. Now in Sydney, I founded Aurai Sensing to bring that stack to horticulture: self-powered field nodes that read soil and canopy and decide on-device.
此后做过采摘机器人、智能车视觉、FPGA 加速 CNN、管道泄漏溯源机器人，以及面向农业的 Sub-1GHz 无线传感网络，过程都完整记录在我的 Lab 里。现在我在悉尼创立了 Aurai Sensing，把这套技术栈带进园艺农业：自供能田间节点持续读取土壤与冠层数据，并在端侧做出决策。
My workflow is agent-native. Coding agents drive toolchains, firmware, CAD and data pipelines, while I stay close to what the hardware is actually doing: the oscilloscope, the power trace, the field.
我的工作方式是 AI-Native 的：编码 Agent 驱动工具链、固件、CAD 与数据管线，而我专注于硬件的真实反馈——示波器、功耗曲线与田间现场。
wenqi@lab:~$ whoami --verbose name Wenqi Wang role Physical-AI engineer · Founder, Aurai Sensing base Sydney, Australia trained Measurement & Control Tech · Instruments loop sense → infer on-device → act silicon STM32 · CC1310 · ESP32 · MSPM0 · Jetson agents Codex · Claude Code · n8n · Activepieces status ● building first field pilot, Sep 2026 wenqi@lab:~$
03 / Journey
China Jiliang University
Optoelectronic-detection stream of a Washington Accord–accredited programme, ranked 15 of 254 nationally (ARWU 2024). First project: an STM32F103 minimum-system board I designed in Altium and soldered myself.
《华盛顿协议》互认体系认证专业，2024 软科专业排名 15/254。第一个项目：用 Altium Designer 自主设计并焊接调试 STM32F103 最小系统板。
Competitions学科竞赛
2nd prize, National Embedded Chip & System Design Competition (East, 2023). 3rd prize, National Optoelectronic Design Competition (East, 2023). Provincial prizes in electronic design (TI Cup), mechanical design and physics innovation, and 1st prize in the university’s “Internet+” innovation contest.
2023 全国大学生嵌入式芯片与系统设计竞赛东部赛区二等奖；2023 全国大学生光电设计竞赛东部赛区三等奖；TI 杯省级电子设计竞赛、浙江省大学生机械设计竞赛、物理实验与科技创新竞赛等省级奖项；校级“互联网+”创新创业大赛一等奖。
National Innovation Training Program国家级大学生创新训练计划
OpenMV and K210 vision guiding an STM32F407-driven arm with a soft gripper: line following, PID tracking and a full SolidWorks assembly. Concluded 2024.
OpenMV 与 K210 视觉引导 STM32F407 控制的机械臂和柔性夹爪，实现巡线、PID 跟踪与目标抓取；SolidWorks 完成整机装配，2024 年结题。
Industry placements企业实践
Summer practice at a 3D-scanning company and a precision-metrology company, seeing measurement science at production scale. The same year: a registered software copyright for an image-processing microscope coal-dust detector.
在 3D 扫描与精密测量企业的暑期实践，让测控专业方向变得具体；同年获得《基于图像处理的显微镜下煤粉检测软件》计算机软件著作权。
Graduation projects毕业设计
Designed and hand-soldered CC1310 RF boards with a localised BOM, and built a Raspberry Pi 4B + ESP32 system that combines environment sensing, image recognition and remote control.
自主设计并焊接 CC1310 射频板（元器件国产化替代），并基于树莓派 4B 与 ESP32 搭建集环境监测、图像识别与远程控制于一体的系统。
The University of Sydney
Moved to Sydney for postgraduate engineering: communication networks, technology management, and a capstone that became a company’s first deployment.
来到悉尼攻读工程研究生：通信网络、技术管理，以及一个最终成为公司首个落地部署的 Capstone 项目。
Aurai Sensing Pty Ltd
IoT- and AI-driven precision irrigation for high-value horticulture: solar field nodes, a 915 MHz self-healing mesh, Field OS, and an agent growers can ask questions.
面向高价值园艺的 IoT + AI 精准灌溉：太阳能田间节点、915 MHz 自愈 Mesh 网络、Field OS，以及种植者可以直接提问的农业 Agent。
University of Sydney Capstone悉尼大学 Capstone 项目
A full node mesh running continuously across a real horticultural property.
完整的节点 Mesh 在真实园艺农场中持续运行。
04 / Selected work
Sixteen projects from 2021 to now. Each “Build log” link opens the full write-up in my Lab (in Chinese). 2021 年至今的 16 个项目，点击「实验记录」可在 Lab 中阅读完整过程。
Founder · 2026 →
Self-powered field nodes read soil moisture at three depths, canopy temperature and VPD every 15 minutes. They decide on-device and run regulated deficit irrigation block by block.
自供能田间节点每 15 分钟读取三个深度的土壤湿度、冠层温度与 VPD，在端侧决策，并按地块执行调亏灌溉（RDI）。
auraisensing.com.au
Platform · 2026
Gateway registration with unique IDs, device ownership, MQTT heartbeats and live online state, a time-series store and dashboards. LLMs read the sensor data and explain what it means.
网关注册与唯一 ID、设备归属、MQTT 心跳与在线状态、时序数据库与可视化面板，并让大模型直接分析传感器数据、给出结论。
Read the build log阅读实验记录Fed BOMs and module specs to a frontier model, iterated a modular housing in SolidWorks, then 3D-printed prototypes to check holes and fit.
把元器件清单与模块规格交给大模型，在 SolidWorks 中迭代模块化外壳，再 3D 打印打样验证孔位与结构。
Build log实验记录At an n8n hackathon: opened the speaker, wired UART, and let an agent guide the firmware backup and SSH access, working toward voice → LLM → n8n automations.
n8n 黑客松：拆机、接 UART 串口，在 Agent 指导下完成固件备份与 SSH 登录，打通「语音入口 + LLM + n8n 自动化」。
Build log实验记录Pre-registered architecture and training experiments for 12-class speech-command recognition on TI’s MSPM0G5187 hardware NPU, measured on the board itself.
在 TI MSPM0G5187 硬件 NPU 上做 12 类语音命令识别，按预先登记的方案进行结构搜索与训练，并在板上实测。
Research in progress研究进行中A 3D scan of me (30,104 vertices, 14 morph targets) rendered live in WebGL. The same bust fronts my Lab, and it’s the one at the top of this page.
我的三维扫描半身像——30,104 个顶点、14 个 Morph Target——在 WebGL 中实时渲染，也是本页顶部的那个模型。
Back to the bust回到模型My own schematic and PCB for a CC1310 RF node with a localised BOM, hot-plate soldering of tiny packages, and radio link tests.
自主设计 CC1310 射频节点原理图与 PCB，元器件国产化替代，热板焊接微小封装并完成通信测试。
Build log实验记录A Raspberry Pi 4B + ESP32 platform that combines environment monitoring, image recognition and remote control for radish microgreens.
基于树莓派 4B 与 ESP32，集环境监测、图像识别与远程控制于一体的萝卜苗种植系统。
Build log实验记录A national innovation project: OpenMV line following, K210 target detection, and an STM32F407 arm with a soft gripper.
国家级大创项目：OpenMV 巡线、K210 目标检测，STM32F407 控制机械臂与柔性夹爪完成抓取。
Build log实验记录An integrated harvest system with a transport cart, a basket-swap mechanism, first-pass inspection and automatic box changing.
集运送小车、换篮机构、初级检测与自动换箱于一体的采收系统。
Build log实验记录A YOLOv5 pipeline that catches scratches, pits, cracks and rust on steel, replacing fatigue-prone manual inspection.
用 YOLOv5 检测钢材划痕、凹坑、裂纹与锈蚀，替代易疲劳的人工目检。
Build log实验记录A competition build covering model conversion, pruning and quantisation, IP-core calls, and ARM A9 + FPGA heterogeneous inference on a remote edge-AI platform.
竞赛项目：模型转换、剪枝量化、IP 核调用，在远程边缘 AI 平台上实现 ARM A9 + FPGA 异构推理。
Build log实验记录TSP route planning, a mecanum-wheel car and a robot arm, plus quantised classifiers squeezed onto an MCU camera (OpenART).
TSP 路径规划、麦克纳姆轮小车与机械臂，以及部署在 MCU 摄像头（OpenART）上的量化分类模型。
Build log实验记录Pneumatic inchworm propulsion with GPS, LiDAR and OpenMV vision, streaming to the cloud to build a digital twin of the pipe.
气囊链式推进，结合 GPS、激光雷达与 OpenMV 视觉，数据上云构建管道数字孪生。
Build log实验记录LLaMA3 running in Docker on a laptop GPU, with retrieval over a personal knowledge base and live web-API calls.
在笔记本 GPU 上用 Docker 运行 LLaMA3，结合个人知识库检索（RAG）与实时 Web API 调用。
Build log实验记录STM32 low-level control under ROS on a Jetson Nano, with LiDAR mapping, a depth camera and obstacle avoidance.
STM32 下位机 + Jetson Nano 上的 ROS：激光雷达建图、深度相机与自主避障。
Build log实验记录05 / What’s next
Three directions I keep investing in, each one built on something that already runs. 三条持续投入的方向，每一条都建立在已经跑起来的系统之上。
06 / From the Lab
The full notebook lives on WordPress: build logs, project write-ups and experiments, mostly in Chinese. 完整的笔记在 WordPress 上：实验记录、项目复盘与学习过程。
lab.auraisensing.com.au
07 / Connect
Open to collaborations in Physical AI, AIoT and agent-native hardware, and to conversations with growers, researchers and fellow builders. 欢迎在 Physical AI、AIoT 与 Agent 驱动的硬件开发方向合作，也欢迎种植者、研究者和同行交流。