MY FIRST-YEAR EXPERIENCE AT IIT-D AI HACKATHON

Hey everyone! Recently I got selected in the round 2 of Sprint4Good AI Hackathon, a 12 hours hackathon conducted by IIT Delhi. So get ready to witness some really energising vibes, tasty refreshments and of course, tons of insights!
In the first round, we (me and my friend) were asked to submit our ideas after selected one of their three topics. After a lot of brainstorming, we decided to develop an Exam-oriented AI learning platform. We submitted it with our fingers crossed, hoping out idea would get recognised.
And it did! On my way home from an exhausting Wednesday at college, I got an email from the organising team about our selection. We were thrilled! The second round, which 
was the final round, was going to be conducted in IIT Delhi campus. But our happiness was short-lived as we realised we would be surrounded by some incredibly tough competitors. So the only chance to advance further was to start early.

By Thursday, we had decided what tech-stack we needed to use - Streamlit for frontend, Ollama as our LLM, Python as our main language (obviously) and MongoDB as our database. It was the first time we were using MongoDB and it took some time to get it going.

By Friday, we had finally got hold of the MONGO_URI and we connected our main file (app.py) with the MongoDB. We were pleased from our progress and convinced we had an edge in comparison to our competitors.

Soon, the day of hackathon came. We decided to commute via metro. After changing a few lines, we reached IIT Delhi's own metro station exactly on time. As we got into the campus, we were soon itching to click some photos of the campus. After a few quick snaps, we directly went into the venue. We were distributed ID-cards and we eventually sat down. As the power supply and wifi were being connected to our laptops, we were served breakfast. After gobbling it down and hearing the introduction speech, we got into work.

What surprised us the most was the fact that the hall included people from almost all corners of the world. People had travelled from places like Chennai to participate in this hackathon. We knew it was going to be tough for us. For starters, we were a team of only 2 members. All the teams had at least 3 members. Also, we both were first-years. Our lack of experience in participating in such events was surely a big drawback in our chances of coming to the top. But we decided to give it a try.

We decided to use Streamlit because our primary goal was to make our platform deliver very efficient results. Hence, we did not have enough time to code the whole frontend as we were too engrossed in the backend. We used Ollama gemma3:4b model for this project which turned out to be quite efficient. We did some coding (and some vibe coding) to almost build the whole website from the scratch before the lunch arrived. After completing the thali, we did some minor debugging and went to the mentors for suggestion.

Before I tell you what happened with the mentor, I want to tell you know what features we managed to include in our Exam-oriented AI learning platform. First, the user is asked to import a PDF file containing your notes. Then you need to click a button called Ingest PDF, the application would break your notes into chunks, embed them and store them in MongoDB. After that you have three features- Ask questions, Summarise and Quiz.

Ask Question: This feature allows the student to ask any question from the notes. Now this is significant because searching from Google or GenAI platform would give you answers which are not always aligned to your syllabus. This platform answers questions using only your notes and identifies the page number where the answer was found. If the answer is not available within the notes, it will inform you. In fact, you can also communicate in 'Hinglish'.

Summarise: Gives a very detailed and brief summary of the notes so that every topic is covered in a concentrated piece of writing.
Quiz: This is a cool feature which we had added. If you want, you can straight up generate 10 MCQ questions from the notes and see your live score as you do the quiz or you can also upload your PYQs and generate MCQs whose question are aligned to the PYQs you have uploaded. The quiz also has Easy, Medium and Hard levels so that you can challenge yourself if you want.

When the mentor saw, he asked us to compare our response (in ask question feature) with that of ChatGPT. It turned out that our response was better in comparison which pleased the mentor.

After rehearsing our presentation in front of the mentors for the elimination round, we were ready. Even though we gave our best, there were lot more others who had build some really cool stuffs. By the looks of the products they made, it was easy to figure out that they were building them for months. The 'edge' we had prepared really did not help us a lot. Nevertheless, we got an opportunity to talk to few other participants and we exchanged our ideas. When the top 6 teams presented the ideas in front of the jury (we were not part of the top 6 selected teams), we got to know a lot of things and got heavily inspired. When we slowly passed the main gate of IIT-D (not before clicking the iconic IIT-D photo), we return our homes with some really unique ideas.

Even though we lost, we gained something more valuable. Experience and new rush of ideas.