They Turned Down $550K. Here's What That Actually

Varun Vummadi grew up in a small town in Andhra Pradesh. His parents were government teachers. When he ground his way into IIT — one of the most competitive engineering programs in the world — and then landed a $550,000 job offer from a leading quantitative trading firm in New York, that wasn't just a good outcome. For his family, that was the outcome. The whole point of the grind.

He turned it down anyway.

His co-founder, who by Varun's account ranked third in their entire IIT cohort, was in a similar position — reportedly fielding what Varun describes as one of the highest job offers available to a graduating engineer in India at the time, though neither the firm nor the specific figure is independently verified. They'd known each other since freshman year. They said no to the sure thing together and applied to Y Combinator instead, three days before Varun was supposed to show up for work.

I've talked to a lot of people who've made bets like this — not $550K bets, but bets where the math didn't add up and they did it anyway. The baker who walked away from a hospital administrator job to open a shop on a corner where two previous bakeries had failed. The machinist who turned down a plant manager promotion to open his own shop because he was tired of building other people's balance sheets. What Varun's story has in common with theirs isn't the dollar amount. It's the conversation with the parents. "My dad was super mad," Varun says simply. "It's kind of like a big fight thing in home." You can feel how much is compressed into that sentence.

The Interview That Almost Wasn't

What happened at the YC interview is worth understanding, because it shaped everything that followed.

Varun had prepared obsessively — the idea, the market size, the pitch. The interviewers, led by YC partner Harj Taggar, didn't ask about any of it. They looked at what Varun and his co-founder had actually built and done — serious research experience in machine learning, Kaggle competition wins, deep technical chops — and essentially said: your idea is wrong, but you're not. Pick something else.

"I thought genuinely the interview went so horrible that we are not going to get in," Varun recalls.

They got in.

This is the part of the YC mythology that gets glossed over in the highlight reels: the acceptance wasn't for the business plan. It was a bet on two people. Harj Taggar connected them with the Coursera COO and a roster of ed-tech veterans who confirmed what he already suspected — the education technology space was a dead end for them. Within a month of joining the batch, Varun and his co-founder pivoted entirely.

Three Pivots, One Customer Signal

What they built next wasn't the customer support product either. It was a tool for making AI language models cheaper and faster to run — what's called "fine-tuning," essentially training a smaller, more efficient model to do what a large expensive one does, at a fraction of the cost. They open-sourced their models, topped the rankings on Hugging Face (a platform where researchers benchmark AI model performance), and on the strength of that traction, reportedly raised a $4 million seed round — a figure Varun mentions but that I'd note comes from his own account and hasn't been independently confirmed by a public filing or announcement.

The fine-tuning business turned out to have a ceiling. Cheap AI processing is useful, but it's not a destination — it's infrastructure, and infrastructure plays tend to get absorbed by bigger players or commoditized fast. What Varun noticed, watching his actual paying customers, was that two use cases kept growing: coding assistance and customer support. His customers were telling him something that his original roadmap hadn't predicted.

This is a story I've seen play out in businesses with a lot less runway and a lot less technical sophistication. I once talked to a woman who opened an Italian restaurant and spent two years trying to figure out why her most popular nights weren't tied to her best dishes or her seasonal menu changes. Eventually she figured out it was because she was one of the few places in her neighborhood that was genuinely fun on a Friday — good noise level, good bar, tables that encouraged staying. She thought she was selling Italian food. She was selling Friday night. She stopped fighting the signal and leaned into it. Varun did the same thing: stopped insisting on the business he'd planned and built toward the one his customers were already voting for with their wallets.

Eight People, One Big Contract

The customer support AI product works like this: instead of calling a company, waiting on hold, getting routed through a robotic phone tree, and eventually reaching a human who reads from a script, you get an AI that can actually resolve your issue — return a package, troubleshoot a device, change a flight. According to Giga's own figures (unverified by independent parties), traditional automated phone systems resolve somewhere between 10 and 15 percent of calls without human intervention. Giga says their system handles 60 to 70 percent, with a stated goal of reaching 90 to 95 percent. Put plainly: if you're running a customer service operation with twelve people fielding calls, that ratio matters enormously. You're not replacing all twelve people — but you're fundamentally changing what those twelve people spend their time doing.

Their first major customer was Zepto, a fast-scaling Indian delivery company. The second was DoorDash — and that one requires some context.

When Giga went after DoorDash, they were a team of eight people competing against what Varun describes as a 400-person, well-funded company. DoorDash, for its part, is a major company that takes its infrastructure seriously. Eight people should not win that contract.

They won.

How? Part of it was a warm introduction — a YC partner connected Varun to someone at DoorDash, and YC-to-YC trust is a real currency in that world. Part of it was a three-month pilot where Giga simply didn't go down and hit their numbers. But the introduction only gets you in the room. What kept them there was the product performing.

Varun is honest about one aspect of this: when they signed Zepto and started pursuing DoorDash, they weren't fully aware of competitors like Sierra — an AI customer support company with a prominent founding team including former Google VP Clay Bavor and former Salesforce co-CEO Bret Taylor, and considerably more capital behind it. "We don't know Sierra and [others] existed when we signed Zepto," he says. The timeline of when competitive awareness set in is likely more complex than a single conversation can capture, but his broader point stands: they weren't running from a competitor. They were running toward a customer.

"Our entire product mentality is: are the customers willing to pay you? Can you deliver a lot of value to them?" he says. "That was a pretty much stronger mentality against competition."

The AI Company That Runs on AI

Giga's internal operations are worth a separate conversation. They operate on a company value Varun describes without irony as "automate, automate, automate." Their sales team pulls call transcripts and runs competitive analysis through AI tools. Scheduling, meeting notes, configuration changes — all automated. Most strikingly, Varun estimates that without AI coding tools, they'd need six or seven times as many engineers as they currently employ. He's careful to say it's not only about cost. "It's better without context switching. It's better for you to own the thing and build the entire thing rather than having a lot of people working on it."

Their engineering interviews reflect this. They have candidates write code using AI tools — and then they take the AI away and ask the candidate to explain and modify what was just built. The point isn't to trick anyone. It's to answer a specific question: do you actually understand what you built, or did you just prompt your way to something that looked right?

That's a trust question, not a technical one. And it's the same question every small business owner asks when they hire someone — can you do this when things get weird, or can you only do it when everything's going smoothly?

What Varun Would Tell His Former Self

The advice Varun gives to the students in the YC India audience is the same advice I've heard from people who've built businesses on much thinner margins in much less forgiving markets: stop falling in love with your idea and start finding out if anyone will pay for it.

"It's never about the idea. It's about if somebody is willing to pay you money for it," he says. "Is somebody willing to pay real money if you solve the problem for the value that you delivered?"

He also says something that I think gets at the real reason the $550K rejection mattered: "Burning the boats is a good thing to start. Things get really real if you burn the boats. That's when I really felt it — because when the company was not working, me and my co-founder were thinking, 'We rejected all these job offers. What are we going to do?' It actually forces you to make things."

The bet Varun and his co-founder made wasn't really about money. It was about whether they wanted to find out what they were capable of. The $550K wasn't the cost of starting the company. It was the cost of finding out the answer.

Most people never pay that price. Which is exactly why most people never get the answer.

— Dorothy "Dot" Williams, Small Business & Entrepreneurship Correspondent, Buzzrag