10 Final-Year IoT Project Ideas for Engineering Students in Assam

A good final-year IoT project solves a problem you can point at, works reliably in front of an examiner, and shows data it collected over time. Below are ten ideas built around problems that are real in Assam (floods, monsoon humidity, farms and patchy networks), each with the core parts and the one detail that lifts it above a copied tutorial.

Most of these come from things I've built myself, from a voice-controlled plant companion to a gesture-controlled robotic arm. Where an idea is a design I haven't built, I say so.

The ten ideas at a glance

1. Voice-controlled plant monitor

A plant pot that hears a wake word, reads its sensors and answers out loud, all on the device with no cloud call. This is the SmartPot build: an ESP32-S3 runs a TensorFlow Lite wake-word model, FreeRTOS juggles four tasks (audio, sensors, MQTT and a small web server), and a DFPlayer Mini plays the spoken reply.

What makes it stand out: on-device speech is rare in student projects. Measure the time from wake word to reply and put that number in your report. If you want the background, I wrote up running TensorFlow Lite Micro on the ESP32-S3.

2. Rain-aware smart irrigation

Most irrigation projects switch a pump on when the soil is dry. Mine, the garden automation system, only waters when four conditions pass: soil moisture is below a threshold, it hasn't rained in the last 6 hours, no rain is forecast in the next 2 hours, and the temperature is in a safe range.

What makes it stand out: the decision logic. In a state with a long monsoon, not watering before rain is the useful part. Use a capacitive soil probe, not the cheap resistive kind that corrodes.

3. A weather station that checks itself

My live weather system paired a BME280 (temperature, humidity and pressure) with OpenWeatherMap data and compared the two to spot sensor drift.

What makes it stand out: most weather stations just display numbers. Comparing your sensor against an outside source, and flagging when they disagree, shows you understand that sensors are not always right, especially in damp air.

4. Gesture-controlled robotic arm

A webcam watches your hand, MediaPipe finds 21 landmarks per frame, and Python maps them to six servo angles that stream to an ESP32 over WebSockets. My 6-DOF arm used a PCA9685 driver for six MG996R servos and reached under 50 ms from gesture to motion on local Wi-Fi.

What makes it stand out: it demos well. Budget for a proper power supply, because six large servos draw far more current than a USB port can give, and add a filter to smooth the jitter.

5. Remote feeder with jam detection

A 3D-printed auger driven by a servo, controlled from a web dashboard with schedules and a feed log. In my fish feeder, an IR sensor confirmed that food actually dropped, the firmware aborted the cycle if it jammed, and the maximum portion was enforced in firmware, not just the app.

What makes it stand out: the fail-safes. Explaining why the limit lives in firmware is exactly the kind of design reasoning a viva rewards.

6. River-level flood alert node

I haven't built this one; it's a design sketch. Floods hit Assam almost every monsoon: in 2024 one ASDMA bulletin counted 26 districts flooded and nearly 14 lakh people affected. A node with an ultrasonic or radar distance sensor mounted over a stream or drain, sending readings over GSM, can raise an SMS alert when the level rises fast.

What makes it stand out: alert on the rate of rise, not just a fixed level, and design for power cuts with a solar panel and battery. Be clear in your report that it's a student prototype, not a replacement for official warnings.

7. Audio-reactive modular light panels

My Nanoleaf-style panels are 3D-printed hexagons with WS2812B LEDs, touch sensors on the edges, and an ESP32 that runs an FFT on audio to drive 12 lighting modes at 60 fps.

What makes it stand out: it mixes signal processing, embedded firmware and mechanical design in one project, which is useful if your team has members from different branches.

8. Classroom air-quality monitor

Not something I've built, but simple and useful. Measure PM2.5 and CO2 in a classroom or hostel room, and show the reading against India's National Air Quality Index bands (Good, Satisfactory, Moderate and so on) rather than raw numbers.

What makes it stand out: log a week of data and show when CO2 climbs during lectures. A real pattern from your own campus is worth more than a perfect circuit diagram.

9. Humidity logger for a tea nursery or greenhouse

Several nodes spread over a field send temperature and humidity over LoRa to one gateway that has Wi-Fi or GSM. I wrote about how I'd design IoT for a tea garden, including which sensors are worth it and which aren't.

What makes it stand out: range testing. Walk the field, record signal strength at each point and put the map in your report. Check that your LoRa module is configured for the band allowed in India.

10. Room energy monitor

A non-invasive current clamp around one wire of an appliance circuit, an ESP32 reading it, and a dashboard showing daily usage. Also a design idea rather than one of my builds.

What makes it stand out: calibrate it against a known load and state your error. Never open a mains board yourself; have a qualified electrician fit the clamp.

Picking the right board

For nine of these ten, a regular ESP32 is enough. Pick an ESP32-S3 if you're running a model on the device, as in idea 1. If you're unsure, I compared the ESP32 and the Raspberry Pi Pico separately.

What I'd do

If I were choosing again as a final-year student, I'd pick one idea, not three, and spend the extra time on reliability. From my own builds, these are the things that made the difference:

• Design for the real constraint. SmartPot answers without the internet; the irrigation system checks for rain before watering. Pick one constraint and make the project about it.
• Add a fail-safe. A jam detector, a portion limit, an alert when a sensor drifts. It shows you thought about what goes wrong.
• Log data early. Get a breadboard version sending readings in the first few weeks, so you have days of real data by the end.
• Plan for humidity. In Assam, an outdoor node without a sealed enclosure and conformal-coated board may not survive a monsoon.
• Write down what didn't work. A short "limitations" section is honest, and it's often the most interesting part of the viva.

If you're a student or college lab in Assam and want help with firmware or a stuck prototype, here's how I work on IoT projects.