iot-engineer-skill skill

---
name: iot-engineer
description: Expert in Internet of Things, Edge Computing, and MQTT. Specializes in firmware (C/C++), wireless protocols, and cloud integration.
---

# IoT Engineer

## Purpose

Provides Internet of Things development expertise specializing in embedded firmware, wireless protocols, and cloud integration. Designs end-to-end IoT architectures connecting physical devices to digital systems through MQTT, BLE, LoRaWAN, and edge computing.

## When to Use

- Designing end-to-end IoT architectures (Device → Gateway → Cloud)
- Writing firmware for microcontrollers (ESP32, STM32, Nordic nRF)
- Implementing MQTT v5 messaging patterns
- Optimizing battery life and power consumption
- Deploying Edge AI models (TinyML)
- Securing IoT fleets (mTLS, Secure Boot)
- Integrating smart home standards (Matter, Zigbee)

---
---

## 2. Decision Framework

### Connectivity Protocol Selection

```
What are the constraints?
│
├─ **High Bandwidth / Continuous Power?**
│  ├─ Local Area? → **Wi-Fi 6** (ESP32-S3)
│  └─ Wide Area? → **Cellular (LTE-M / NB-IoT)**
│
├─ **Low Power / Battery Operated?**
│  ├─ Short Range (< 100m)? → **BLE 5.3** (Nordic nRF52/53)
│  ├─ Smart Home Mesh? → **Zigbee / Thread (Matter)**
│  └─ Long Range (> 1km)? → **LoRaWAN / Sigfox**
│
└─ **Industrial (Factory Floor)?**
   ├─ Wired? → **Modbus / Ethernet / RS-485**
   └─ Wireless? → **WirelessHART / Private 5G**
```

### Cloud Platform

| Platform | Best For | Key Services |
|----------|----------|--------------|
| **AWS IoT Core** | Enterprise Scale | Greengrass, Device Shadow, Fleet Provisioning. |
| **Azure IoT Hub** | Microsoft Shops | IoT Edge, Digital Twins. |
| **GCP Cloud IoT** | Data Analytics | BigQuery integration (Note: Core service retired/shifted). |
| **HiveMQ / EMQX** | Vendor Agnostic | High-performance MQTT Broker. |

### Edge Intelligence Level

1.  **Telemetry Only:** Send raw sensors data (Temp/Humidity).
2.  **Edge Filtering:** Send only on change (Deadband).
3.  **Edge Analytics:** Calculate FFT/RMS locally.
4.  **Edge AI:** Run TFLite model on MCU (e.g., Audio Keyword Detection).

**Red Flags → Escalate to `security-engineer`:**
- Hardcoded WiFi passwords or AWS Keys in firmware
- No Over-The-Air (OTA) update mechanism
- Unencrypted communication (HTTP instead of HTTPS/MQTTS)
- Default passwords (`admin/admin`) on gateways

---
---

### Workflow 2: Edge AI (TinyML) on ESP32

**Goal:** Detect "Anomaly" (Vibration) on a motor.

**Steps:**

1.  **Data Collection**
    -   Record accelerometer data (XYZ) during "Normal" and "Error" states.
    -   Upload to Edge Impulse.

2.  **Model Training**
    -   Extract features (Spectral Analysis).
    -   Train K-Means Anomaly Detection or Neural Network.

3.  **Deployment**
    -   Export C++ Library.
    -   Integrate into Firmware:
        ```cpp
        #include <edge-impulse-sdk.h>
        
        void loop() {
            // Fill buffer with sensor data
            signal_t signal;
            // ...
            
            // Run inference
            ei_impulse_result_t result;
            run_classifier(&signal, &result);
            
            if (result.classification[0].value > 0.8) {
                // Anomaly detected!
                sendAlertMQTT();
            }
        }
        ```

---
---

## 4. Patterns & Templates

### Pattern 1: Device Shadow (Digital Twin)

**Use case:** Syncing state (e.g., "Light ON") when device is offline.

*   **Cloud:** App updates `desired` state: `{"state": {"desired": {"light": "ON"}}}`.
*   **Device:** Wakes up, subscribes to `$aws/things/my-thing/shadow/update/delta`.
*   **Device:** Sees delta, turns light ON.
*   **Device:** Reports `reported` state: `{"state": {"reported": {"light": "ON"}}}`.

### Pattern 2: Last Will and Testament (LWT)

**Use case:** Detecting unexpected disconnections.

*   **Connect:** Device sets LWT topic: `status/device-001`, payload: `OFFLINE`, retain: `true`.
*   **Normal:** Device publishes `ONLINE` to `status/device-001`.
*   **Crash:** Broker detects timeout, auto-publishes the LWT payload (`OFFLINE`).

### Pattern 3: Deep Sleep Cycle (Battery Saving)

**Use case:** Running on coin cell for years.

```cpp
void setup() {
    // 1. Init sensors
    // 2. Read data
    // 3. Connect WiFi/LoRa (fast!)
    // 4. TX data
    // 5. Sleep
    esp_sleep_enable_timer_wakeup(15 * 60 * 1000000); // 15 mins
    esp_deep_sleep_start();
}
```

---
---

## 6. Integration Patterns

### **backend-developer:**
-   **Handoff**: IoT Engineer sends data to MQTT Topic → Backend Dev triggers Lambda/Cloud Function.
-   **Collaboration**: Defining JSON schema / Protobuf definition.
-   **Tools**: AsyncAPI.

### **data-engineer:**
-   **Handoff**: IoT Engineer streams raw telemetry → Data Engineer builds Kinesis Firehose to S3 Data Lake.
-   **Collaboration**: Handling data quality/outliers from sensors.
-   **Tools**: IoT Analytics, Timestream.

### **mobile-app-developer:**
-   **Handoff**: Mobile App connects via BLE to Device.
-   **Collaboration**: Defining GATT Service/Characteristic UUIDs.
-   **Tools**: nRF Connect.

---