I'm Going to NVIDIA!

Sophomore year has been an insane journey so far, at least compared to me being a bum as a freshman.

I wanted to share some fun stories of the past two quarters, what's happening in Spring, and where I see myself going in the future.

Part 1: Winning Cal Hacks

This all traces back to an exact moment where I made a random decision with an insane butterfly effect.

1. I was at the library doing some homework—doomscrolling a bit—when I saw on Instagram that UCSB ACM had a welcome week event. I was honestly not tryna go, especially since my bum ass was not in the mood to meet people. But then I saw Sean's location was there, so I was like "f– it let's go". (W decision)
2. He left 5 mins after I got there. But this was a good chance to meet some new faces: (1) Aman was another CCS Computing student in my same embedded systems class, so we decided to partner up. (2) Calvin who is SB Hacks sponsorship director, and my eventual Cal Hacks teammate. (3) Ritam was the SB Hacks president, who I met last year during midnight hackathon yoga and in Sid's AI Physics class.
3. I asked Ritam: any advice for getting into research? I was interested in the NLP lab, but I wasn't sure how to approach it. He literally pulled up Slack with a list of new PhD students and told me the secret method—just email the PhDs, not the professor.
4. I emailed Saaket, who did research on Computer-Use Agents (CUA) and built to Simular's open source Agent S repo. I said something (lowk goofy in hindsight) about his paper and that I wanted to collaborate. He responded "sure let's talk sometime"
5. Ritam tells me there's a research BBQ the next day. I wasn't sure about it since it's for PhD students, but I go anyway. Yuh. Saaket is there, so I introduce myself properly in person. We start meeting regularly for a theory of mind project.
6. I join SB Hacks using Ritam nepo powers. During one of the weekly meetings, I realize that Surya and Ishan applied too late for Cal Hacks and didn't get in. So I ask around—Calvin also needed a teammate too.
7. We go to Kaiju Ramen and talk about ideas. I realized that remote desktop software sucks and it would be nice to "text your computer," and have a CUA agent do stuff for you. This idea popped up because (1) I studied CUA papers since I had to do my initial email to Saaket and (2) because AWS went down and I had to bike all the way back home to get the homework PDF for my class with Aman.
8. We decided this idea was cooler than last year's winner, and that we left the restaurant thinking we had a real shot of winning.
9. We won.

So really, it all started by deciding to go to that random ACM event. Nice!

Part 2: NVIDIA Referral + Research Project that Perfectly Fit

After winning Cal Hacks, I had some things like Amazon Annapurna Labs and SpaceX, but it didn't lead anywhere.

Apple also reached out, but I submitted the application before Cal Hacks so maybe I would've gotten it anyway. I also failed that (was too nervous about sticking to a pre-recited "tell me about yourself").

Luckily, during winter break Josh mentioned he could ask his brother to refer me to NVIDIA. Especially since Cal Hacks is a cool plus.

Two months later, on a random Tuesday I got an email: "NVIDIA - Machine Learning Engineering Intern - Summer 2026"

Yuhhh.

And the email mentioned RL, post-training, and Nemotron. Literally the dream. Though when I showed Surya he said it's probably more data generation side. We shall see I guess.

I think what stood out most was my research project that I was working on in the NLP lab, which was a benchmark for embodied theory of mind in AI agents. I built a scalable data generating method to scale difficulty and diversity to create an evolving benchmark that should always be challenging for frontier agents. I think they saw that as a good fit for what they wanted.

Startups Offers

At that time, I was also very lucky to get an offer from startups affiliated with UCSB NLP Group: ChipAgents (William Wang is the CEO) and Simular (Eric Wang is head of research, was impressed that I used Agent-S in Cal Hacks).

I moved ChipAgents to the Spring quarter, and Simular is TBD. Possibly Fall 2026 quarter, but the offer may not stand that long or (fingers crossed) Ritam can slide me w/ a calm Cohere research internship.

The only issue was that in this month-long period, I was scrambling to manage offer deadlines between companies and I was very stressed out by this (not to mention my 7 class btw, yes this is a quarter system so 28 units).

NVIDIA Interview Process

The interview went quite well I think. My first one was just a chat about general LLM training. He basically just said: "okay you want to train the next GPT model, what do you do?"

I stumbled a bit at first but luckily I remembered some stuff from class, especially since I just had a lab on DPO/RLHF which was very helpful to remember the equation and tradeoffs of each. Afterwards I had some time to ask some questions so I asked about applying RLMs and some other data gen techniques that they used.

The second and final round interview was a coding round. At first it was just a Jupyter notebook with some Scikit-Learn/Pandas task, but I didn't know how to use it. So luckily we were able to switch to a more traditional Leetcode-style probleml, which I was able to solve.

Overall, was super chill I think.

Part 3: Offer and What's Next

The first time the recruiter called me, I was in the middle of an Elden Ring Neightreign boss fight, so I lowk thought it was a spam call and ignored it. That caused a lot of stress when I tried calling back 3 times without her picking up!

But finally, after an hour-long walk to Sands and back, she finally called back. I got the offer!

To celebrate, I made a fun video using AI (this is a joke okay!).

Conclusion

Looking forward to this summer. Only sad thing is no free food or gym... which are some perks I will miss from ChipAgents.

And I think one of the senior research scientists is joining UCSB as a professor, might be worth hitting him up for as an additional collaborator.