5 AI Project Ideas for College Students and PhD Students in Assam

Assam is growing fast, and research in the area of AI is also rapidly increasing. Here are some AI areas where college students as well as PhD students can put their thoughts into and potentially build something useful for Assam.

People are already working on Assamese-based AI, where AI can understand, talk and reply in the Assamese language.

But to become a differentiator, you cannot just use another AI API, send it Assamese text to get a reply, and then process the output to generate a voice response. These things are already being done by the big players. We are essentially just putting together things that have already been created, and anyone with a good understanding of computer science can create a basic version of it in a few days.

But here is the opportunity.

See, everything that LLMs are producing comes from what they have been trained and fine-tuned with. And although Assamese is already represented in some language datasets and AI projects, there is still a huge opportunity to build much richer, Assam-specific datasets.

So if you and your team can somehow get legal access to what Assam's libraries, archives, universities and other institutions can provide — whether it is a book from a small author, historical documents, literature, folklore or even material around traditional and cultural practices — it could be extremely valuable.

Collect the data as much as possible, but always make sure you have the proper rights and permissions to use it.

Then clean it, structure it and train or fine-tune a model specifically to understand Assam's language, culture and context.

That would be a killer project.

Go ahead if your team can do it, because Assam really needs that.

Data is where everything lies.

Another very important topic, as floods hit Assam every year.

Try to build an AI/ML model that understands flood patterns, so that it can predict the possibility of flooding and potentially estimate how severe or widespread it could be.

To build something like this, you will need the right data. This could include historical satellite images along with information such as rainfall, river levels, terrain and previous flood data.

Getting access to this data can be one of the first challenges. Try reaching out to your professors and research departments and ask whether they can help you access suitable government or research datasets. There are already government and research systems using satellite data for flood monitoring in Assam, so the opportunity is to build on top of existing data and research rather than starting completely from scratch.

Once you have the data, you can start training your model. For example, you could provide it with previous satellite images and manually label which areas were flooded and when. By combining this with other historical data, the model can start learning the patterns associated with flooding.

This would be much more of a computer vision, geospatial and machine-learning problem than an LLM problem, and that is actually what makes it interesting.

Once the model has been trained and tested properly, you could connect it with newer or live data and see whether it can predict the possibility and extent of flooding.

Of course, this would need a lot of testing before it could ever be trusted for real-world flood warnings, but as a research project, this could be a very interesting area to work on.

Yes, we already have many projects that can detect number plates and check whether a vehicle has proper documents or not.

But this AI would be something different.

In this project, you can actually try to detect whether someone is breaking road laws, such as crossing a divider illegally, jumping a traffic signal, driving on the wrong side of the road, illegal parking and many other kinds of violations.

Computer vision can already detect and track vehicles, but the difficult part is understanding what is actually happening on the road.

For example, simply detecting a vehicle near a divider does not mean it has crossed it illegally. The AI needs to understand the road, the traffic signal, the vehicle's movement and the context around it.

So this cannot directly go into production. It would need proper datasets, testing, validation and clear rules for how the AI should identify a wrongful action.

But as a research project, this would again be a great place to start and experiment.

The world has seen what floods have done to people living in areas of Assam, including the Sivasagar region, and every year there are people who come forward to help.

But sometimes, even those who want to help cannot reach certain areas because they are difficult to navigate and extremely isolated. Roads can be underwater, villages can become cut off, and some houses and families can be left behind without receiving the basic necessities they need.

Helicopters are one way of delivering supplies, but they are not always practical or available for every situation. And many people who want to help may themselves be afraid of entering deep or fast-moving floodwater.

So, can a small unmanned RC and AI-assisted inflatable boat be of some help?

A properly designed boat could potentially carry a useful payload if the motor, battery, hull and overall weight are calculated correctly. It could be operated using a remote control and a live camera feed, allowing the operator to navigate it without physically entering the flooded area.

And this is where AI can make things even more interesting.

The boat could monitor its battery level, GPS location and communication signal. If the signal becomes weak, instead of simply continuing forward, it could switch to a predefined safety mode or attempt to return to its starting point.

With further research, AI-assisted navigation and obstacle avoidance could also be explored.

The idea is not to immediately build a fully autonomous boat and send it into a real flood.

Start small.

Build the RC boat.

Test the payload.

Test the battery.

Test the camera.

Test the communication range.

Then start adding AI-assisted features.

If everything works properly, such a system could eventually become a useful tool for delivering essential supplies to people who are difficult to reach during a flood.

AI-based drone detection and tracking can be very useful for the police, defence and critical infrastructure, not just in Assam but across the country.

We have already seen how drones are changing modern conflicts, and detecting small drones can be difficult, especially when they are flying at a distance or in challenging conditions.

So instead of trying to build another generic drone, one interesting research direction would be an AI-based system that can detect, identify and track potentially dangerous or unauthorised drones.

The system could use cameras and computer vision to identify a drone, estimate its movement and raise an alert when it enters a restricted area.

Research in this field is already going on, but there are still many areas where the technology can be improved — especially detection at longer distances, poor visibility, different drone shapes, false detections and tracking.

A student or research team could focus on the detection and tracking side first and build a prototype that can identify and follow drones in a controlled environment.

That alone would be a very challenging and useful AI research project.

Keep following us to get more project ideas like this, as we bring new ideas on a regular basis.

Our team loves to brainstorm, and while doing that, we come across many ideas. We have decided to keep adding some of these ideas to our website so that they do not just remain ideas.

Not all of them will turn into real projects.

But maybe some will.

And if even a few of these ideas can be turned into something useful for Assam, then we all could benefit from it.