All projects
IoT + ML

FreeSpot

Real-time IoT classroom availability (Final Year Project)

Shows students and staff, live on their phones, whether each room is free, a class is running, a teacher is late, or a class was cancelled or moved.

PythonESP32mmWave radarMQTTFastAPIPostgreSQLReact PWAscikit-learnDocker Compose

307 ms

p95 event-to-dashboard

lab measurement

234

backend tests

0

cameras or microphones

The problem

Students waste time walking between rooms to find out whether a class is on. Cameras would solve it but invade privacy.

Architecture

  1. ESP32 + radar + RFID
  2. MQTT broker
  3. FastAPI status engine
  4. PostgreSQL
  5. React PWA (live)

How it works

  • ESP32 nodes with privacy-friendly mmWave radar (presence only — no camera, no microphone) and RFID teacher check-in.
  • MQTT with per-room credentials streams events to a FastAPI status engine that combines sensors, timetable and device health.
  • Mobile-first React PWA with live Server-Sent Events updates.
  • Random Forest occupancy model benchmarked against a calibrated threshold rule with session-wise cross-validation, with automatic fallback.

Safety & reliability

  • An offline sensor is never shown as a free room.
  • A class is never declared cancelled from sensor data alone.

What I learned

Writing the status engine as a pure function with no clock or database inside made every safety rule testable on a laptop.