Why I'm Leaving Industry for Grad School

It's 2 AM and I'm sitting in my PG in Bangalore, staring at a half-finished GRE practice set and a cold cup of chai. My roommate is asleep. The Airtel WiFi is being its usual unreliable self. And I keep asking myself the same question I've been asking for months: am I really doing this?

Yeah. I'm really doing this.

The Decision

I'm leaving a good ML engineering job to pursue an MS in Computer Science in the US. On paper, this doesn't make a lot of sense. I was earning well for Bangalore, working on interesting problems at Myelin Foundry, and had the kind of practical experience that companies value. My parents were happy. My bank account was growing. Things were stable.

But here's the thing. Two years of production ML showed me exactly how much I don't know. I could build and deploy models, sure. But when I read papers from CVPR or NeurIPS, I'd hit walls. The math felt shaky. The theoretical foundations had gaps. I could implement architectures but I couldn't always reason about why they worked or predict how they'd behave on new problems.

That gap started bothering me more and more.

The Conversations

I had a lot of late night conversations about this with friends and my roommate. The ones who went to the US for grad school a year or two before me painted an honest picture. It's expensive. The visa situation is uncertain. You go from earning money to burning through savings. The first semester is brutal.

But they also said something that stuck with me: "You come back thinking differently." Not just more knowledge, but a different way of approaching problems. Research thinking. The ability to read a paper and actually critique it instead of just implementing it.

My parents took some convincing. My mom's concern was practical: why leave a paying job to take on debt and uncertainty? My dad was more philosophical about it but I could tell he was worried too. The H-1B lottery, the cost of living, being far from home. These aren't small things.

Honestly, I get their concerns. I share them.

The GRE Grind

Preparing for the GRE while working full-time was one of the more painful experiences of my life. I'd come home from work, eat whatever Swiggy delivered, and study vocab words for two hours. Quant was fine because the math is relatively straightforward. Verbal was torture. I spent weekends doing practice tests and trying to convince myself that memorizing obscure English words was a reasonable use of my time.

I ended up with a decent score. Not perfect, but good enough for the programs I was targeting.

What Actually Pushed Me Over

There was a specific moment, I think. I was debugging a deployment issue at Myelin, something with model quantization causing accuracy drops on specific input patterns. I eventually fixed it, but the process felt mechanical. I knew the what but not the deep why. I could pattern-match from Stack Overflow and GitHub issues but I couldn't derive the solution from first principles.

I want to be the person who understands the first principles. Not just someone who uses the tools, but someone who could build them.

Looking Forward

I'm scared, honestly. I'm trading a comfortable life in Bangalore for uncertainty in a new country. I'll be the oldest person in some of my classes. I'll be broke for a while. I might hate it.

But the friends who went before me keep saying the same thing: "I wish I'd done it sooner." And the version of me that stays in Bangalore, comfortable and well-paid but increasingly frustrated by the ceiling, that version scares me more than the uncertainty.

So here I am. Applications submitted. Fingers crossed. Chai getting cold. Let's see what happens.