Level: Beginner | Time: 3-4 hours | Cost: $40-80

Build a distributed sensor network with multiple edge nodes collecting environmental data and reporting to a central gateway. Learn edge data collection, MQTT messaging, and basic IoT architecture.

Hardware Needed:

• 3x ESP32 or ESP8266 microcontrollers ($5-10 each)
• 3x DHT22 temperature/humidity sensors ($5 each)
• 1x Raspberry Pi 4 as gateway ($35-75)
• Power supplies (USB chargers or battery packs)
• Breadboards and jumper wires
• MicroSD card for Raspberry Pi

Instructions:

1. Setup Gateway: Install Mosquitto MQTT broker on Raspberry Pi
2. Configure Nodes: Program ESP32s to read sensors and publish to MQTT
3. Test Connectivity: Verify each node can connect to WiFi and broker
4. Implement Data Flow: Nodes publish sensor readings every 10 seconds
5. Create Dashboard: Use Node-RED on Pi to visualize data from all nodes
6. Add Edge Logic: Program nodes to only send data when values change significantly
7. Monitor System: Track message frequency and node uptime

What You'll Learn:

• MQTT publish-subscribe messaging pattern
• Distributed sensor network architecture
• Edge preprocessing to reduce network traffic
• Gateway pattern for aggregating edge devices

Extensions:

Add different sensor types, implement data logging to time-series database (InfluxDB), create alerts for threshold violations, or add battery monitoring and deep sleep modes for power efficiency.

Level: Intermediate | Time: 4-6 hours | Cost: $150-250

Deploy a computer vision model on edge hardware to detect objects in real-time without cloud connectivity. Learn model optimization, inference at the edge, and performance tuning.

Hardware Needed:

• Raspberry Pi 4 (4GB+ RAM recommended) or NVIDIA Jetson Nano
• Pi Camera Module or USB webcam
• Optional: Google Coral USB Accelerator ($60) for faster inference
• Power supply and cooling (heatsink/fan)
• MicroSD card (32GB+)

Instructions:

1. Setup Environment: Install TensorFlow Lite or PyTorch on edge device
2. Choose Model: Download pre-trained YOLOv8 or MobileNet model
3. Optimize Model: Convert to TFLite format and apply quantization
4. Test Inference: Run model on sample images, measure FPS and latency
5. Stream Video: Connect camera and run real-time object detection
6. Add Logic: Only save frames or send alerts when specific objects detected
7. Performance Tuning: Adjust resolution, model size, and frame rate

What You'll Learn:

• Deploying ML models to resource-constrained devices
• Model quantization and optimization techniques
• Real-time inference performance considerations
• Edge vs cloud tradeoffs for computer vision

Extensions:

Train custom model with Edge Impulse or Roboflow, implement object tracking across frames, add privacy features (blur faces, local-only processing), or create multi-camera distributed vision system.

Level: Intermediate | Time: 4-5 hours | Cost: $50-100 + cloud costs

Design and implement a complete data pipeline from edge sensors through gateway to cloud storage and analytics. Learn data flow design, edge aggregation, and cloud integration patterns.

System Components:

• Edge sensors (multiple ESP32s or Arduino IoT devices)
• Edge gateway (Raspberry Pi running local processing)
• Cloud platform (AWS IoT Core, Azure IoT Hub, or Google Cloud IoT)
• Data storage (cloud database or time-series DB)
• Visualization dashboard (cloud-based or self-hosted)

Architecture Design:

1. Map Data Flow: Diagram data path from sensors to cloud to dashboard
2. Define Edge Processing: What happens at edge vs in cloud?
3. Select Protocols: MQTT for edge-to-gateway, HTTPS for gateway-to-cloud
4. Plan Storage: Time-series data, metadata, and alert history
5. Design for Failure: What happens when cloud is unreachable?
6. Implement System: Build each component and integrate
7. Test Scenarios: Normal operation, network loss, sensor failure

What You'll Learn:

• End-to-end IoT system architecture
• Edge aggregation and buffering strategies
• Cloud service integration (IoT platforms)
• Fault tolerance and offline operation

Extensions:

Add OTA (over-the-air) firmware updates, implement edge caching with automatic cloud sync, create alerting system with SMS/email, or add device management and monitoring dashboard.

Level: Intermediate-Advanced | Time: 5-8 hours | Cost: $30-60

Train and deploy a machine learning model on a microcontroller to classify sounds. Learn TinyML workflow from data collection through deployment on ultra-low-power devices.

Hardware Needed:

• Arduino Nano 33 BLE Sense (has built-in microphone) or
• ESP32 with I2S microphone module
• USB cable for programming
• Computer for training (Edge Impulse Studio is cloud-based)

Instructions:

1. Define Problem: Choose 3-5 sounds to classify (e.g., voice, clapping, knocking, silence)
2. Collect Data: Record 2-3 minutes of each sound class using Edge Impulse
3. Design Features: Extract audio features (MFE, MFCC) suitable for classification
4. Train Model: Use Edge Impulse to train neural network classifier
5. Test Performance: Evaluate accuracy and confusion matrix
6. Optimize for Edge: Quantize model to reduce size and memory usage
7. Deploy to Device: Export Arduino library and flash to microcontroller
8. Test Real-time: Verify classification works in real-world conditions

What You'll Learn:

• Complete TinyML workflow from data to deployment
• Audio feature extraction for embedded ML
• Model optimization for microcontrollers
• Real-time inference on ultra-low-power devices

Applications:

Voice command recognition, industrial sound monitoring (detecting machine faults), baby monitor with cry detection, smart home with gesture sound recognition, or wildlife monitoring.

Time: 2-3 hours | Materials: Spreadsheet, paper for diagrams

Create a decision framework for determining when to use edge computing versus cloud-only architecture for different application scenarios.

Framework Factors:

• Latency Requirements: Real-time (<10ms), interactive (<100ms), or batch (>1s)
• Bandwidth Costs: Volume of data, upload vs processing locally
• Privacy/Compliance: Sensitive data that can't leave premises
• Reliability: Need to function during network outages
• Scale: Number of devices, geographic distribution
• Compute Requirements: Processing complexity, hardware acceleration

Exercise Scenarios:

1. Retail store with 20 cameras for customer flow analysis
2. Factory with 500 vibration sensors for predictive maintenance
3. Smart home with voice assistant and automation
4. Autonomous drone fleet for agriculture monitoring
5. Hospital patient monitoring with vital signs

For each scenario, analyze requirements and recommend edge vs cloud architecture with justification.

Time: 2-3 hours | Materials: Drawing tools (digital or paper)

Design network topology for a multi-location edge deployment considering data flow, redundancy, and communication patterns.

Design Scenario:

Chain of 10 retail stores, each with 15 IoT devices (cameras, sensors, beacons). Stores connect to regional data centers for aggregation before cloud upload.

Design Requirements:

• In-Store Layer: Device-to-gateway connectivity within store
• Store-to-Regional: Aggregation and local processing at region
• Regional-to-Cloud: Selective upload of insights and alerts
• Redundancy: Failover paths if primary connection lost
• Management: How to monitor and update all edge nodes

Deliverables:

Network topology diagram showing all layers, data flow diagram indicating what data moves where, failure mode analysis, and bandwidth/compute requirements at each level.

Time: 2-4 hours | Materials: Spreadsheet, cloud pricing calculators

Perform detailed cost comparison between edge and cloud architectures for a specific use case over 3-year period.

Cost Analysis Scenario:

Video analytics system for 50 cameras recording 24/7. Compare two architectures:

• Cloud-Only: All video streams to cloud, processing happens centrally
• Edge-Hybrid: Object detection at edge, only events/metadata sent to cloud

Cost Factors:

• Hardware: Edge devices, servers, cameras, networking equipment
• Connectivity: Internet bandwidth costs (especially upload)
• Cloud Services: Compute, storage, data transfer, ML inference
• Maintenance: Updates, monitoring, support labor
• Power: Electricity costs for edge devices

Analysis:

Create 3-year total cost of ownership model for both architectures. Consider different scenarios: 10 cameras, 50 cameras, 200 cameras. Identify break-even points and key cost drivers.

Team Size: 4-6 people | Time: 3-4 hours

Design a smart building system using edge computing for energy optimization, occupancy monitoring, and environmental control.

Team Roles:

• IoT Architect: Overall system design and device selection
• Edge Developer: Local processing and gateway logic
• Data Engineer: Data pipeline and storage design
• Security Lead: Network security and data privacy
• Facilities Manager: Building operations perspective

Design Challenge:

10-floor office building with 500 rooms. Design IoT system for:

• Occupancy sensing and lighting automation
• HVAC optimization based on room usage
• Energy monitoring per floor
• Air quality and environmental monitoring
• Fault detection for building systems

Deliverables:

System architecture diagram, device and sensor specifications, edge processing logic description, data flow and storage plan, security model, and 3-year budget estimate.

Team Size: 3-5 people | Time: 2-3 hours

Create implementation plan for industrial edge computing system for predictive maintenance in manufacturing facility.

Scenario:

Manufacturing plant with 50 production machines. Implement vibration and temperature monitoring for predictive maintenance using edge computing.

Planning Requirements:

• Sensor Selection: Types, placement, communication protocol
• Edge Infrastructure: Gateways, compute capacity, networking
• ML Models: Anomaly detection, remaining useful life prediction
• Integration: Connection to existing SCADA/MES systems
• Rollout Plan: Pilot, phased deployment, training
• ROI Justification: Reduced downtime, maintenance cost savings

Deliverable:

6-month implementation roadmap with milestones, risk assessment, resource requirements, and success metrics.

Team Size: 4-6 people | Time: 2-3 hours

Plan the physical deployment of edge devices across distributed locations considering logistics, configuration, and ongoing management.

Deployment Scenario:

Deploy 200 edge devices across 50 retail locations in 10 cities. Each location needs 4 devices (cameras + edge gateway).

Logistics Challenges:

• Procurement: Ordering hardware, managing inventory
• Pre-Configuration: How much setup before shipping?
• Shipping: Coordinating delivery to locations
• Installation: DIY with instructions vs professional installers?
• Commissioning: Network configuration, testing, activation
• Ongoing Management: Updates, monitoring, troubleshooting
• Support: Help desk, on-site support, replacement devices

Deliverable:

Deployment playbook with timeline, checklists for each phase, training materials, and support escalation plan.