Developed a full-stack quiz platform with distinct user and admin
functionalities.
Implemented user authentication, role-based access control, and
session management using Flask Designed relational database
schemas to optimize Object-Relational Mapping (ORM) operations.
Built
dynamic, server-rendered pages using the Jinja2 template
engine.
Added features like quiz management, score
tracking, search/filter options, and interactive data
visualizations using Chart.js.
Built an AI-powered resume evaluation platform with dynamic
scoring and competitive leaderboard.
Integrated LLM APIs using Mira Flows Network by writing a Flow
that takes resume data as input and provides detailed
feedback/review.
Implemented strict scoring constraints within the Flow, with
results reflected on the Leaderboard.
December 2024
The big one. In-between the lines, we got selected for the SIH Finale. Our problem statement demanded of us to develop a program that denoises PSR images of Lunar surfaces as captured by the OHRC camera on-board Chandrayaan-II. We developed a CNN model based on UNet architecture trained over a custom-made dataset of 600 PSR images. Essentially we took 600 sunlit images and ruined them with noise until they matched the characteristics of normalized PSRs. Using the synthetically damaged PSRs and clean PSRs as noisy-clean pairs, we trained our model, cooking a (borrowed) RTX 4060 for 4 hours, hoping it works on the actual PSR images. And it did, splendidly.
At the halfway point of 4am, I socialized with our wonderful opponents. I'd have graciously lied to them and undersold our progress if there was any. I didn't need to, we were all mutually screwed. At 7am, I entered the "python script04.py" command for the 400th time. The Fourier Transform to destripe the images worked. At this moment, my brain could stop iterating "Experience matters, win does not". To hell with that, we had a fighting chance. We had the chance to win India's biggest goddamn Hackathon. All we needed to do was to improve our presentation!
So did we manage to improve our presentation skills? Most certainly not.
But did we win SIH 2024? I think you know...
October 2024
Attended my first Open Source Event by OSW. Met a lot of wonderful people and learnt a lot about FOsS, Git, elasticsearch and GenAi.
But more importantly, I got cosmetics for my laptop!
September 2024
This takes us to TicTechToe '24, my first big-scale Hackathon hosted by DA-IICT University. I was the FrontEnd dev for my team simply because I had no other skills. Tho I'd proudly like to tell I wrote the backend for the login and signup pages of our project. Only took 48 hours.
More proudly, I'd like to share that my team was selected as the Top 15 Finalists out of a total of 150 teams present at the venue.
August 2024
This is the day of SIH internals, where we developed a mini-project that takes an image captured by any Lunar orbitter and creates a 3D terrain for it. It is a sad attempt at Procedural Terrain Generation for Lunar terrain before we knew that is an actual thing. You can see one of it's outputs being rendered as the background of this webpage (if you have a CPU greater than Intel Pentium).
This is also the day I realized plotting 2.7 million nodes without using Numba in Python is generally not a good idea :')
December 2023
If my life were a Bell Graph, what you're witnessing here... is the global maxima of it all. What comes next is arguably better or worse, based on your priorities. This is the first and last instance of my late night academic grinds.
May 2023
If there exists a time where I'm more clueless than I'm right now, it is this. I was juggling between Architecture and Engineering and I didn't wish to pick either. But then I heard a calling, for me, through the clouds...
"Pick CompSci...pick CompSci...pick CompSci..." and so I did.
...it was my father's voice.
CS 1 O(1) Society
Thanks for reaching out. I'll get back to you soon.