College Recommendation System

How do you choose between a better college with a less preferred branch and a preferred branch at a different college?

We built a recommendation system around this question during VCET Hackathon Reloaded 2021, developing the algorithm, backend, UI, and overall product from scratch in 30 hours.

Our project, Ukku, went on to win second prize in the finale.

The Problem

For students choosing an engineering college, the decision isn't simply about finding the highest-ranked college.

There is often a trade-off between:

• College
• Branch
• Academic performance
• Personal preferences
• Career goals
• Other factors that matter differently to each student

We wanted to build something that treated this as a decision problem, rather than simply returning a list of colleges sorted by a single score.

What We Built

The system takes a student's academic profile and preferences and uses them to generate relevant college and branch combinations.

The core idea was to model the trade-offs between college and branch rather than assuming that one is always more important than the other.

The project included:

• A recommendation algorithm
• College and course data processing
• Preference-based matching
• Backend APIs
• A web interface for exploring recommendations
• The overall product and business approach

Everything was designed and implemented by our team during the hackathon.

Recommendation Approach

Instead of treating recommendation as a simple ranking problem, we wanted the system to expose the factors behind the recommendation.

A student's preferences could change the relative importance of different options.

For example, a student might prefer:

A highly ranked college with a less preferred branch

while another might prefer:

A preferred branch at a different college.

The recommendation system was designed around this trade-off.

The goal wasn't to tell students which option was objectively correct, but to narrow down the possibilities and give them a more useful starting point for making their own decision.

Built in 30 Hours

The entire project was developed during VCET Hackathon Reloaded 2021.

In roughly 30 hours, we went from the initial problem statement to a working product, including:

• Understanding the problem
• Designing the recommendation logic
• Building the backend
• Developing the UI
• Integrating the system
• Testing the complete flow
• Preparing the final presentation

That constraint made the project as much an exercise in prioritization and product thinking as it was in implementation.

What I Learned

The biggest lesson was that a recommendation system doesn't necessarily need to produce a single "correct" answer.

For problems involving personal preferences, the more useful system is often one that:

• Makes its assumptions clear
• Exposes relevant trade-offs
• Narrows down a large search space
• Gives users enough context to evaluate the results themselves

That distinction between ranking something and helping someone make a decision has stuck with me beyond this project.

The Team

I built this project with:

• Jaideep More
• Keshav Mishra
• Vedant Kokate

It was one of those projects where the time constraint forced us to learn quickly, divide responsibilities, and make decisions without over-engineering.

Recognition

Our solution won Second Prize at VCET Hackathon Reloaded 2021.

The hackathon was organized by Vidyavardhini's College of Engineering and Technology.

Tech

Python · Machine Learning · Recommendation Systems · Backend · Web Development

Hackathon

VCET Hackathon Reloaded 2021

Built from scratch in approximately 30 hours.

Result: Second Prize

Links

• Github Repo