Stop Wasting Time: 8 Secrets About GSoC Nobody Tells You

Stop Wasting Time: 8 Secrets About GSoC Nobody Tells You

First of all, Who am I and Why should you give a f**k

Hello, My name is Shuvam Pal and yes I cracked Google Summer of Code in my very first year of engineering. I achieved in my first year what most people only dream of achieving in their four years of college 😈.

Just kidding, I’m no prodigy, just another 19-year-old kid who stills Googles how to not mess up his Git treeā˜ŗļø. I am still trying to digest the legacy left behind by my seniors at Point Blank.

Here’s how it all started…

The Discord Meet

It was the month of December, and there was this announcement about a discord meet being held to talk about GSoC. I had heard only titbits before about seniors cracking GSoC, but didn't really know what it really was. So just curious I joined the meet. By the end of the meet it was clear to me what it exactly was and how big cracking GSoC was. Also one thing I concluded was that nobody had cracked it in their first year in our college yet.

I thought to myself, "Hmm, what if I could be the guy cracking it in first year?". It was just one amusing thought back then but anyways, reluctantly, I got to work.

I started looking for Organisations to contribute to. I shortlisted two: FreeBSD, which seemed really cool, and InVesalius, which matched my tech stack and wasn’t as crowded. After a lot of trial and error I managed to run FreeBSD locally, but contributing to it would’ve been a nightmare (and I’ll admit, I was a bit lazy). So I went with InVesalius.

The First PR

Reading and understanding this codebase was an arduous task. I had written Python programs before, but nothing close to what powered InVesalius. Nevertheless, I spent the whole month trying to figure things out and fix issues. Along the way I found a few bugs of my own, fixed them, and raised PRs.

By the end of the month I had a handful of PRs but still no response from the maintainers. I started to get worried. Then, on the very last day of the month, I got a notification: one of my PRs was merged. That feeling? Unmatched. Getting your first PR merged feels absolutely awesome.

The Proposal

I continued finding issues and raising PRs. I slacked off sometimes in between, but kept the progress going somehow. I also started contributing to another Org called IOOS. At this point I was making Open Source contributions just for the fun of it and also I was getting to learn a lot. By mid March I had made quite some progress and now it was time to choose a project and start writing a proposal. I had no clue what to do but nevertheless I chose the hardest project I could find in the list.

Writing the Proposal was another nightmare. Either I was overlooking things or making them overcomplicated. I took feedback from the mentors and went through a total of four iterations before I settled on the final version (I rewrote my whole proposal right before the submission day :) ).

I let out a huge sigh of relief after clicking the submit button. The very next day, my fellow PB members and I set off on a trip to Coorg, a much needed refreshment for everyone who had been working on GSoC. That trip was one to remember. The perfect relaxation plan. Whenever I listen to this song, it takes me back to those nights.

There is so much pain in the world, but not in this moment.

The Aftermath

Back from Coorg, I continued contributing and communicating with the maintainers. I tried focusing on other things and learning new skills for a while. I don't know where I got it from, but I carried this quiet confidence that the results would turn out in my favour.

Before I knew it, a month had passed, and D-day had arrived. Nerves were high, everyone was hoping for the best, fingers crossed. Some people got their rejection mails, others received their acceptances. It was chaos, a swirl of mixed emotions.

Meanwhile, my inbox was still empty. And then - BAM! - at 11:48 p.m. on May 8th, a notification popped up. I opened it, and there it was: 'Welcome to Google Summer of Code 2025!'.

I can’t fully describe the rush of emotions that hit me in that moment. It was pure happiness and pride. I could hardly sleep that night, my head buzzing with a thousand thoughts.

But cracking GSoC was only half the story. The real challenge was about to begin…

Coding Begins!

Except...it didn't for me. After the results were announced, I slacked off quite a bit. I wanted to spend time on other things and explore different topics. I felt confident because I knew I had most of the project figured out while writing my proposal (or so I thought).

I didn’t actually start building the project seriously until late June. I worked on it piece by piece, and at first it felt like a cakewalk. But once the components came together, the real challenge became clear. The complexity of the implementation and the sheer scale of the project only hit me after I had finished fleshing out the pipeline.

The Encounter With Ephialtes

I was bombarded with thousands of errors, and quality issues in the output. Then I tried converting the models to ONNX and my GSoC literally turned into a nightmare. The issue was way beyond the expertise I had in ML so far. I egged AI with questions to no avail. Finally I found the answer lying in one corner of a remote blog post. It didn't really fix the issue but was an efficient workaround.

I continued grinding. Then just when I thought I had it running successfully, I saw the output I was getting and was horrified. It was ugly. I could have submitted it, it met the basic requirements, but after so much work, I couldn’t settle for anything less than the best result.

My mistake? Relying too much on AI. I learned my lesson the hard way. All I was left with after weeks of prompting was, NOTHING! Finally I thought - fuck AI. I pulled an all-nighter, manually changing and testing every little detail to trace the root cause, and by the morning I had it all working, The final product was right there in front of me, and trust me, it was beautiful.

My mentors tested it out, and I even went ahead to design and ship two additional products. At last, after a wrecked lower back, several sleepless nights and countless empty packets of instant coffee, it was done. I had pulled through. I had really clutched it this time.

Mentor Feedback

I hope you got some value out of this read, and you don't make the same mistakes as me.

Despite what people say about this being a great feat, I know how much I still have to grow. At the end of the day I am just another average coder trying to find my place in the world.

The horrors persist but so must we.

Here are my links: Github | Linkedin (drop in a follow maybe :) )

And that’s a wrap! Finis coronat opus.