Qiuzi Huang — AI Agent Engineer

The short version

Send this config to the AI you already use — Claude, Cursor and others. It reads my projects, history and skills straight from this site, so you do not have to dig through the page.

{
  "mcpServers": {
    "hqz": {
      "url": "https://sundahqz.com/hqz/mcp"
    }
  }
}

On Claude Code it is one command: claude mcp add --transport http hqz https://sundahqz.com/hqz/mcp

Then just ask it:

Which ESP32 projects has Qiuzi Huang built?
What has he done on AI agent deployment?
What were the hardest technical problems he solved?

Timeline

Most recent first. Work and education.

2025.04 — 2026.11
Shenzhen Lianyu Technology Co., Ltd.

AI Hardware Development Manager

Led debugging, development and mass production of AI edge devices — core boards, sensor integration, compute scheduling and stability; worked with algorithm and software teams on model deployment and performance tuning.
- Edge AI
- Core boards
- Sensors
- Compute scheduling
- Mass production
2023 — 2025
Shenzhen Oukaipusi Energy Technology Co., Ltd.

Energy Storage Product Manager

Developed residential and commercial & industrial energy storage products; drove feature improvements and post-sales support, with customer satisfaction above 95%.
- Residential ESS
- C&I ESS
- Product development
- Support
2017 — 2023
Shenzhen Xingnengda Technology Co., Ltd.

Li-ion BMS & Home Storage Engineer

Built BMS hardware and control algorithms, extending battery life by 20%; took a residential storage product from concept to mass production, lifting market share by 10%.
- BMS
- Hardware
- Algorithms
- Mass production
2014 — 2016
Dongguan Fengbang New Energy Co., Ltd.

R&D Engineer

Optimised lithium battery production processes, raising yield by 15%; tuned production equipment to cut unplanned downtime by 30%; ran multi-dimensional sample testing.
- Process
- Yield
- Equipment
- Testing
2010 — 2014
Xinhua College of Sun Yat-sen University

B.Eng. Biomedical Engineering

Bachelor of Engineering in Biomedical Engineering, graduated 2014.
- Biomedical engineering
- B.Eng.

Projects

Sized by how much is actually written up — the deeper the notes, the more room the project gets.

AI Agents & Dev Tools

dsh-plugin

Deployment layer for an internal agent platform: auth and permissions, ten in-house plugins, model gateway, session archiving, desktop client distribution — with install, backup and migration tooling.

dsh listens only on 127.0.0.1; caddy terminates HTTPS in front of it, so the agent’s remote code execution stays on the machine and is never exposed to the network. The installer runs front to back — preflight, engine pull, dependencies, preset validation, credential and start-script generation — and is idempotent, so it can be re-run safely.

cli-anything-erpnext

A local ERPNext v16 stack plus a Python CLI that exposes the ERP to agents through the real Frappe REST API — JSON output, dry-run, role-scoped commands.

A local ERPNext v16 Docker stack (MariaDB, Redis, worker, scheduler) plus a Python CLI. The CLI lets agents drive the ERP through real Frappe HTTP APIs, with JSON output, dry-run support, generic DocType operations and role-scoped flow commands.

seeed-cli

A Go CLI that orchestrates AI across an engineering workflow — 13 capabilities, and it syncs skills across four major AI tool directories.

Security scanning, quality checks, architecture review, AI-facing project documentation, daily reports, commit messages and API docs chained so a single pass covers them — and it syncs skills across AI tool directories following the SKILL.md convention.

AI_SKILLS

A collection of 25+ agent skills: embedded platform conventions, Chinese legal research, PDF parsing, and planning / debugging / delivery-verification workflows.

deepseek-harness

Deployment baseline for an open-source agent engine, tracking upstream with a client fork.

Embedded & ESP32

ESP32-BLUETOOTHSPEAKER

A2DP receive → SBC decode → I2S DMA → DAC at 44.1 kHz / 16-bit, 21-step volume, ~130 ms end-to-end latency, with auto-reconnect and shake-to-skip.

An ESP32 Bluetooth A2DP receiver feeding a PCM5102 DAC over I2S at 44.1 kHz / 16-bit stereo, with 21-step quantised volume, pairing that persists across a power cut, and WS2812 plus TFT status showing track and album art.

Welding-Torch-Handle-Screen

Bare-metal APM32F051C8T6 firmware: RS485 to the welder at 19200 bps, a 12-byte binary frame protocol, 120 ms polling, and dirty-flag partial redraw to stop flicker.

Firmware for a welding-torch handle screen on an APM32F051C8T6 (ARM Cortex-M0+, 64KB flash / 8KB RAM). It talks to the welder over RS485 and shows live welding parameters on a 0.96-inch TFT with four-key navigation.

hqz-esp32s3-20250223

Collected ESP32-S3 learning notes and experiments.

bestarc_lvgl_esp32s3

ESP32-S3 touchscreen UI built on LVGL.

esp8266-remote

ESP8266 remote control firmware.

PB03F-BLUETOOTH-mod

BLE module development on PHY62XX.

BESTARC_3_5INCH_remote-control

ESP32-S3 3.5-inch remote: one codebase, two roles, with BLE pairing and factory reset.

ESP-AIFOC

FOC drive for humanoid joint motors: power stage, current sensing, protection, and closed speed/current loops — fixing low-speed jitter and positioning accuracy.

ESP-AI-DOOR

AI smart door with on-device wake and recognition.

Edge AI

k230_ai_camera

A K230 edge inference camera in C (dual-core RISC-V + KPU), reaching above 70% utilisation on typical networks.

An on-device inference camera on Canaan’s K230 (dual-core RISC-V plus KPU), written in C, holding above 70% compute utilisation on typical networks.

jetson-orin-baseboard

Jetson Orin baseboard — upstream open-source project by Antmicro.

01Studio_K230_Resources

Collected K230 development resources (local only, not published).

Other

vibe-pet

A token and task monitor for agents — 15 agents and 14 desktop or hardware pet forms supported.

Turns coding agents — Codex, Cursor, Windsurf and others — into live companions on desktop and hardware. Choose from thousands of characters or make your own in one click; 15 agents and 14 pet forms are supported.

python_test

Python tests and experiments.

Repositories

Own repositories, straight to GitHub.

AI Agents & Dev Tools
hqz-dsh
dsh-plugin
seeed-cli
AI_SKILLS
cli-anything-erpnext
Embedded & ESP32
BESTARC_3_5INCH_remote-control
bestarc_lvgl_esp32s3
esp8266-remote
ESP32-BLUETOOTHSPEAKER
hqz-esp32s3-20250223
ESP-AIFOC
ESP-AI-DOOR
PB03F-BLUETOOTH-mod
Welding-Torch-Handle-Screen-APM32F051C8T6
Edge AI
k230_ai_camera
Other
vibe-pet
python_test

Skills

AI agent engineering

Agent framework plugins & tools
Skills and role presets
MCP and CLI integration
Prompt & context engineering
Multi-agent orchestration
Verification & audit trails

Deployment & ops

Windows / Linux services
Reverse proxy & self-signed PKI
Model gateway & credential governance
Automation (PS 5.1 / 7, Bash)
Backup, migration, DR drills

Application development

Electron desktop apps
Node.js / TypeScript
React web plugins
Document parsing & OCR
FFmpeg media pipelines

Embedded & AI hardware

ESP32 (ESP-IDF)
LVGL
K230 / CanMV edge inference
FOC motor control
BMS & energy storage
Creo / CAD

Contact

Email: sundahqz@gmail.com
Phone: +86 150 1927 3843
GitHub: github.com/hqz-2024

Twelve years of device-side work — compute, power, comms, manufacturability — put to work inside real agent deployments.