Discover how two IIT engineers turned down lucrative jobs to build Giga, an AI customer support platform now serving DoorDash, crypto exchanges, and Fortune ...
How Two IIT Engineers Built a $550K AI Startup: Giga's Customer Support Success
Core Insights
- Turned Down $550K Offers: Two IIT engineers rejected lucrative quant firm positions to pursue their AI startup vision through Y Combinator
- Revolutionary AI Support: Giga increased customer support deflection rates from 10-15% to 60-70%, with goals to reach 90-95% for premium customers
- Major Client Wins: Successfully partnered with DoorDash (competing against a 400-person team with only 8 people), crypto exchanges, and top three telecom providers
- Product Over Sales: The founders discovered that exceptional product quality matters far more than traditional sales tactics in AI-driven companies
- Customer-First Validation: Early payment from customers like Zepto proved product-market fit better than any pitch or presentation
The Rejection That Changed Everything: Y Combinator's Pivotal Role
When Varun, one of the co-founders, walked into his Y Combinator interview, he was terrified. He'd spent weeks preparing, discussing ideas with other YC founders, and crafting the perfect pitch. But Harsh, the interviewer, didn't ask about any of it. Instead of diving into their EdTech startup concept, Harsh simply told them they were "really good engineers" and suggested they work on something else entirely.
"I genuinely thought the interview went horribly and that we wouldn't get in," Varun recalls. Yet this rejection of their original idea became the pivotal moment that led to Giga's existence. Y Combinator's guidance proved instrumental not just for funding, but for connecting them with industry experts who warned them away from the oversaturated EdTech market. Within a month of the batch starting, they'd completely pivoted their entire business direction—a decision that would define their future success.
The timing couldn't have been more dramatic. They were supposed to join a quant firm in three days when they finally received their YC acceptance. This razor-thin margin between two completely different futures illustrates the high-stakes gamble these two engineers were taking. But Y Combinator's network and mentorship transformed what could have been a devastating career decision into the foundation of a transformative company.
From Stanford Research to AI Revolution: The Making of Two Exceptional Founders
Understanding Giga's success requires understanding the founders themselves. Varun came from a small town in Andhra Pradesh where both his parents worked as government teachers. In true Indian middle-class fashion, they expected their son to become an engineer or doctor. He delivered on the engineering part, earning admission to IIT Kharagpur, one of India's most prestigious engineering institutions, where he studied electrical engineering.
His co-founder was equally exceptional—he ranked third in the entire IIT class and maintained nearly perfect grades throughout his college career. The only "B" he received came from a technicality when he accidentally marked attendance for a friend, resulting in a 10-mark penalty. Otherwise, he would have topped the entire campus. This wasn't just academic excellence; it represented an obsessive commitment to mastery that would later define the company's culture.
The two founders took dramatically different paths to success, yet both proved equally valuable to building Giga. While Varun's co-founder epitomized the "book guy"—someone who excels through systematic study and deep understanding—Varun himself was a builder and competitor. During college, he participated extensively in Kaggle competitions, not purely for the intellectual challenge but because winning offered real financial rewards. He earned approximately $50,000 through competition wins, which eventually helped him secure one of those high-frequency trading positions. He was so successful that the platform actually banned him for his competitive dominance.
This combination of academic excellence and practical building experience proved crucial. During his third year at IIT Kharagpur, Varun began researching large language models at Stanford, working on transformer models like BERT during the pre-ChatGPT era. This research experience gave him foundational knowledge of AI systems at a moment when most people didn't understand these technologies existed. His co-founder brought complementary strengths: the ability to think deeply about systems, understand theoretical foundations, and execute flawlessly.
When ChatGPT launched, both founders immediately recognized its potential. Unlike many observers who saw it as a novelty, they understood it as a fundamental shift in what was possible. They wanted to build something on top of it and get into Y Combinator. Varun was reading Paul Graham's essays and watching the 2014 YC Startup School—the classic introduction to startup thinking. His co-founder, knowing him since freshman year, simply agreed: "What could go wrong? We can just apply."
The $550K Decision: Why They Walked Away From Guaranteed Wealth
The choice to start Giga wasn't made in a vacuum. Before their Y Combinator acceptance, Varun received a job offer from one of the leading quantitative trading firms in New York for approximately $550,000 annually. In 2023-2024, this represented an extraordinary opportunity—the kind of paycheck that could set someone up financially for life, especially someone from a middle-class background in India.
His co-founder had similarly impressive offers, including a PhD opportunity from Stanford and what was reportedly one of the highest-paying job offers available in India from a quant firm. By any conventional measure, both founders were positioned to achieve significant financial success through traditional career paths.
Yet they chose differently. This decision created serious family conflict. Varun's father was "initially very upset" when his son pursued the startup path, creating what he describes as "significant conflict at home." His parents, having worked their entire lives as government teachers, had specific expectations for their son's future: stability, predictability, and the kind of social status that came with a prestigious corporate job.
"Coming from a middle-class background, there are significant expectations, especially when turning down a good job offer that would have brought financial stability," Varun explains. "Even I sometimes questioned if I was making the wrong decision."
However, Varun had already internalized Paul Graham's essay on building wealth, which emphasizes building a company and maintaining equity in something significant rather than simply earning a high salary. He showed his parents Y Combinator's videos and success stories of other YC companies, trying to convince them that the risk was worth taking. His core argument: even if the startup failed within a year or two, he could always return to a traditional job. The opportunity cost of not trying was, in his mind, far greater than the risk of failure.
This philosophical stance—that it's better to attempt something exceptional and fail than to accept a comfortable but ordinary path—shaped everything about how Giga was built. When the company wasn't working during its early phases, both founders faced genuine panic. They'd rejected lucrative offers, and failure meant not just personal disappointment but family disappointment as well. This pressure, however, forced them to build better products and make smarter decisions faster than competitors who had financial safety nets.
From Fine-Tuning Models to Customer Support: The Pivot That Defined Giga
Giga's journey from founding to market dominance wasn't a straight line. The company pivoted multiple times before finding its true calling in AI-powered customer support. Understanding these pivots reveals important lessons about startup strategy and product-market fit.
Initially, the founders worked on fine-tuning language models. After getting into Y Combinator, Harsh connected them with the Coursera COO and founders of successful EdTech companies. These conversations were consistently discouraging—every expert told them that building in EdTech was a bad idea. With their original plan invalidated but their Y Combinator batch already started, they needed to pivot quickly.
The turning point came when Varun read a research paper from one of the Databricks co-founders about caching LLMs to reduce costs. At the time, GPT-4 was extraordinarily expensive, and the founders realized that fine-tuning with smaller language models might offer a better economic model. They decided to pursue this direction and open-sourced a bunch of models, eventually topping Hugging Face benchmarks. This achievement generated significant traction and attracted many potential customers reaching out with interest.
"We raised a $4 million seed round," Varun recalls. This was validation that their approach had merit. However, as they worked with customers over the following year, they made a crucial discovery: fine-tuning itself was a poor market for a sustainable business. The only genuine reasons to fine-tune were to reduce costs, improve speed, or meet extreme security requirements. More importantly, they realized fine-tuning sales required treating it as a sales process rather than an engineering process—something neither founder wanted to focus on.
While analyzing their customer base, they noticed something significant: only two use cases were actually growing well among their customers. One was customer support, and the other was coding assistance. They decided to focus entirely on customer support, recognizing it as a market with clear, growing demand and the ability to deliver genuine value.
Zepto, an Indian quick-commerce platform, became their first customer for this new direction. Zepto was scaling rapidly and needed to handle increased customer support volume while maintaining quality. The platform tried Giga's AI agents and became so satisfied that they became the company's first official customer for this service. This wasn't just a customer; it was validation that the pivoted product could deliver real value in a competitive market.
David vs. Goliath: How 8 Engineers Beat a 400-Person Company
The most telling illustration of Giga's product quality came through their competition with DoorDash. This wasn't a hypothetical or a test case—it was a real, head-to-head competition for a major customer's business that revealed something profound about how AI-driven markets actually work.
When DoorDash was evaluating customer support solutions, they had multiple options. The traditional vendor in this space was well-funded, had a team of 400 people, and brought decades of industry experience. Giga, by contrast, was a tiny startup with only 8 people. By every traditional metric—company size, resources, experience, market position—Giga shouldn't have had a chance.
Yet they won. This wasn't luck or politics; it was pure product quality. Varun explains: "We were like eight people going against a 400-person, well-funded company like Doordash, and we won. That's when we truly realized there is a lot of arbitrage in actually building a great product rather than just focusing on a sales team."
The Y Combinator advantage played a role here, but only in opening the door. Carrie (from Y Combinator) introduced them to Tony (at DoorDash). As a YC company, Giga had built-in credibility with DoorDash's leadership, which was itself founded by YC alumni. However, credibility alone doesn't win contracts with companies of DoorDash's caliber. What won the contract was what happened next: a three-month pilot where Giga's system never failed and all performance metrics were excellent.
"DoorDash is a very meritocratic company, and they were willing to trust a small team like ours, which is hard for a company of their size," Varun notes. This trust was earned through demonstrated capability, not promised through marketing materials or impressive presentations. DoorDash saw firsthand that Giga's AI agents worked better than alternatives, delivered measurable business value, and did it reliably.
This DoorDash partnership proved transformative for Giga's trajectory. Once DoorDash—one of the most influential platforms in the gig economy—deployed Giga's technology, other major companies took notice. The partnership served as the ultimate social proof. Suddenly, Fortune 500 companies that might have dismissed a small startup as risky now saw Giga as trusted by one of the world's most successful tech platforms.
The Product Philosophy That Drives Everything: Deflection Rates and Customer Happiness
At its core, Giga's value proposition is straightforward but transformative: replace the traditional customer support workflow with AI agents that customers actually prefer to work with. Traditional customer support typically works like this: call in, get routed to an IVR (interactive voice response system) or basic chatbot, wait on hold, eventually reach a human agent. The best these systems achieve is 10-15% "deflection rate"—the percentage of calls resolved without human intervention.
Giga inverts this model entirely. Their AI agents provide an experience so human-like that customers don't realize they're talking to an AI until told. These agents resolve 60-70% of support issues without any human involvement. For their best customers, they're approaching 90-95% deflection rates. The benefit to customers is profound: you never wait on hold again. You call support, explain your issue, and get it resolved within minutes by an AI that understands your problem contextually and solves it immediately.
This dramatic improvement in deflection rates translates directly to business metrics that companies care about: reduced support costs, improved customer satisfaction (CSAT), higher first-contact resolution rates, and better customer lifetime value. A company like DoorDash or a top cryptocurrency exchange can redirect massive support budgets toward other priorities while simultaneously delivering better customer experiences.
The technical approach Giga developed reveals a sophisticated understanding of what actually matters in deploying AI agents at scale. Rather than building monolithic AI systems, Giga discovered that "it fundamentally boils down to policies, often expressed in markdown files." These markdown files contain the rules, constraints, and procedures that guide how the AI agent behaves in different scenarios.
The iterative improvement process is elegant: improve these markdown-based policy files to positively impact business KPIs. If a support resolution rate is stuck at 30-40%, the question becomes: what policy adjustments would push this toward 90%? This approach applies equally to customer support, compliance workflows, IT service management (ITSM), and IT service desk (ITSD) functions. The same principles that optimize support resolution rate optimization can optimize compliance processes or internal IT workflows.
This insight has opened doors to new customer segments. "We are seeing major consumer companies in the US piloting our solutions for internal support and compliance," Varun notes. The technology isn't just valuable for external customer support; it's transformative for any function that relies on policies, workflows, and human decision-making. This expansion possibilities represents enormous growth potential as companies recognize that policy-based workflows exist throughout their organizations.
The Hiring Philosophy: Recruiting the Exceptional and Building an AI-Augmented Team
As Giga's product philosophy emphasizes building with AI rather than building for AI, this philosophy extends directly into how they hire and structure their team. The founders have deliberately built a company that operates on what they call the "automate, automate, automate" principle—maximum automation wherever possible.
The hiring process itself reflects this philosophy. Rather than traditional interview questions about experience or hypothetical scenarios, Giga's interview process is specifically designed to assess how engineers think about code and problem-solving. They ask candidates to write code with AI assistance available (typically tools like GitHub Copilot or similar), then ask them to modify that code with AI access removed. This two-phase evaluation accomplishes multiple things: it shows whether the candidate understands the code they're writing, not just blindly accepting AI suggestions; it demonstrates how they approach problem-solving without assistance; and it reveals how they actually leverage AI tools in practice.
"We ask them to write code and then remove access to the tool and ask them to change the code without AI," Varun explains. This intentional design evolved as AI models improved. As models get better and better, the question becomes increasingly philosophically interesting: "Do you really need to know how the code works, even if AI does the entire thing?" The team hasn't settled on a single answer, but they're intentionally thinking through these questions rather than reflexively demanding traditional technical knowledge.
Beyond this specific interview methodology, Giga looks for what Varun calls "extraordinary ability and spikiness." They recruit for extreme competence in specific areas rather than general competence across many areas. For Varun, his spike was financial outcomes—he had one of the highest job offers available. His co-founder's spike was academic excellence and systematic mastery, ranking third in IIT while maintaining near-perfect grades. The company actively seeks "the 0.1% of people who would do" specific things exceptionally well.
This hiring philosophy creates a team that can achieve more with fewer people because AI augmentation multiplies what each person can accomplish. If your team members are already exceptional in their domains, and you augment them with AI agents that handle repetitive tasks, eliminate context switching, and automate workflows, you get exponential productivity gains.
"If coding agents didn't exist, we would likely need six to seven times more engineers than our current lean, talented team," Varun states. The math is straightforward: a lean team of exceptional engineers augmented with AI accomplishes what would traditionally require 6-7x the headcount. This isn't just a cost advantage; it's a structural advantage in how fast the team can move and how effectively they can coordinate.
The company uses AI extensively internally. The sales team analyzes Gong transcripts (AI-generated conversation transcripts from customer calls) using AI tools to extract insights about competitor strategies, customer objections, market trends, and sales tactics. Rather than having sales reps manually listen to calls or take notes, AI handles the pattern recognition and insight extraction, leaving sales reps to focus on the strategic and relational aspects of selling.
This approach "has turned many people into builders," Varun notes. Rather than getting bogged down in data processing, analysis, and documentation tasks, team members can focus on building, creating, and shipping. The context-switching costs that plague many organizations essentially disappear because everyone owns complete domains, supported by AI automation rather than fragmented across multiple people and teams.
The Mistake That Became a Lesson: Why Product Beats Sales in AI Companies
Early in building Giga, Varun and his co-founder had a significant disagreement about what mattered most for their company's success. Varun believed that sales should be the primary focus—building a strong sales team, developing sales processes, and executing sales strategies. His co-founder disagreed, emphasizing that product quality should come first.
"I was so wrong in so many ways," Varun now admits. He points to the success of every major AI company as evidence: "If you take a look at all the successful AI companies, it's product. None of them are succeeding because of sales teams. Nobody uses Anthropic for the best sales team."
The contrast is striking. OpenAI and Anthropic, two of the most successful AI companies in the world, don't even compensate their sales teams with commissions. "I don't think they pay. I mean, Anthropic and OpenAI don't even pay sales people commissions. That's how they don't care about sales."
This radical de-prioritization of sales isn't because these companies don't need to sell—they obviously do need revenue. Rather, it reflects a recognition that in AI-driven markets, the product's ability to deliver value is so fundamental that traditional sales tactics become secondary. The question that matters is not "how good is our pitch?" but rather "how good is our product at delivering a lot of value to the customer in a short amount of time?"
"If you can prove that, everything else should follow through," Varun argues. This philosophy guided Giga's strategy throughout their growth. When they proved that their AI agents could deliver dramatically better customer support outcomes than alternatives, sales became easier. When DoorDash saw their system work perfectly for three months in a pilot, the decision to deploy became obvious. When Fortune 500 companies saw DoorDash using the platform, their interest became automatic.
The broader insight extends beyond AI to how modern enterprise software markets work. In an era where customers can evaluate products quickly, where information about capabilities spreads rapidly, and where switching costs are decreasing, the traditional sales-driven model becomes less effective. Companies that build exceptional products find that sales becomes a distribution function rather than a creation function.
However, Varun is careful to note a nuance: this doesn't mean product quality alone is sufficient. "There are a lot of people who will buy your product without a business background," he notes. "You just got to find the right buyer." The key is understanding your Ideal Customer Profile (ICP)—who are the customers most likely to benefit from what you're building and most likely to pay for it?
For Giga, that ICP initially included fast-growing platforms like Zepto and companies with complex customer support needs like DoorDash. Finding the right customers—those with acute problems that your product could solve and budgets to pay for solutions—proved more important than sophisticated sales processes.
From Bangalore to San Francisco: Building in the Right Ecosystem for Deep Tech
As Giga's trajectory accelerated, the founders faced a decision that many Indian founders building deep tech companies encounter: should they build in India or in Silicon Valley? This question has no universal answer, but Varun and his co-founder thought through it carefully.
"We are part of a new generation of Indian-origin founders building companies both in India and San Francisco," Varun notes. "I believe founders should always stay close to their customers, regardless of location. However, for deep tech like generative AI, the Bay Area offers unparalleled access to researchers and innovation."
This represents a pragmatic rather than dogmatic approach. The Bay Area concentration of AI researchers, established AI companies, infrastructure providers, and venture capital creates an ecosystem that's difficult to replicate elsewhere. Access to the latest thinking on language models, transformer architectures, and AI systems implementation is concentrated there. Hiring world-class AI researchers and engineers is easier in the Bay Area because that's where the talent concentrates.
However, this doesn't mean building your entire company in the Bay Area if your customers are elsewhere. "Conversely, if your primary customers are in India, you should build your company here," Varun suggests. This distributed approach—maintaining presence in multiple geographies—is increasingly viable for AI companies. The founders split their time between San Francisco and Bangalore, maintaining connection to customers and talent pools in both locations.
This reflects a broader shift in how global tech companies can operate. With remote work normalized and high-bandwidth communication tools available, the geographic concentration that once limited startup success has loosened. However, for deep technical domains like AI systems, being physically present in innovation hubs provides advantages in hiring, partnership, and staying current with the fastest-moving research.
Future Vision: Automating the Forward-Deployed Engineer and Beyond
As of the conversation, Giga is preparing to launch what the founders believe will be their most transformative product yet: an AI-powered forward-deployed engineer. This role doesn't exist in traditional software companies but has emerged in modern AI-augmented organizations. A forward-deployed engineer combines technical expertise with customer intimacy, working embedded with customer teams to understand their needs, troubleshoot issues, and implement solutions.
The AI version of this role would join customer Slack channels and Google Meets, take notes during meetings, identify patterns in what customers are asking for, and automatically implement policy changes and configure new dashboards based on customer needs. Rather than customers needing to wait for support teams to respond to requests or implement changes, the AI forward-deployed engineer proactively works to solve emerging problems and implement improvements.
"I am very confident that this will revolutionize AI adoption in enterprises," Varun predicts. The insight underlying this confidence is that deployment and adoption are the true bottlenecks in AI's enterprise transformation, not capability. Companies have access to powerful AI models and platforms. What they lack are the resources to properly integrate these tools into their workflows, train teams on effective usage, and continuously refine implementation. An automated forward-deployed engineer would address exactly these bottlenecks.
The vision extends to Giga's internal operations as well. The company's core operating principle is "automate, automate, automate." They're building what they describe as "a generic automation engine that can automate any workflow." Rather than building point solutions for specific automation needs, they're developing foundational infrastructure that can be adapted to automate virtually any process.
The practical examples reveal the scope of this vision. Instead of hiring a personal assistant to schedule meetings, an AI agent does it. Instead of having sales people manually listen to customer calls and extract insights, AI analyzes transcripts and surfaces key insights about competitor mentions, customer pain points, and effective selling strategies. Each of these automations doesn't just save time; it eliminates decision-making and context-switching overhead that slows down human teams.
Key Lessons for Aspiring Founders: What Actually Matters
Reflecting on their journey, Varun emphasizes several lessons that he wishes he'd understood earlier in building Giga. The biggest mistake they made—even after getting into Y Combinator—was working on ideas that generated no revenue for extended periods.
"The biggest mistake we made, even after getting into YC, was working on numerous ideas that generated no revenue or impact for a long time," he admits. "It's not about having many ideas; you can get ten ideas from ChatGPT. The crucial factor is whether someone is willing to pay real money for the problem you solve and the value you deliver."
This emphasis on early customer validation and payment is worth emphasizing. Varun and his co-founder spent time building fine-tuning infrastructure, open-sourcing models, achieving benchmark rankings, and acquiring traction from people interested in the technology. However, none of this converted into sustainable revenue or business viability. Only when they shifted focus to customer support—something actual companies would pay for—did they find product-market fit.
"If they're not, you might be solving a fake problem," he notes. The distinction between a real problem (something customers will pay to solve) and a fake problem (something customers find interesting but won't prioritize their budget toward) is absolutely critical. Many founders spend months or years solving fake problems, building impressive technology for problems nobody cares enough about to pay for.
The approach that proved successful: "This approach of seeking early payment and commitment from customers for new products has been a strong lesson for us." Rather than building in stealth, refining endlessly, and launching to market, they started asking customers to pay early. Zepto became their first paying customer for customer support precisely because Giga asked them to pay for and commit to the product. This forced alignment—the customer has skin in the game, making them honest about whether the product actually solves their problem.
For college students and young entrepreneurs, the biggest takeaway is to stop overthinking and start doing. "The biggest thing is just getting started and trying to sell the thing. Just getting started, jumping in, and burning the boats is a good thing to start."
"Burning the boats" is a reference to Hernán Cortés's historical strategy: when invading a new land, he ordered his ships burned so his soldiers would have no choice but to commit fully to the mission. Applied to startups, it means cutting off fallback options, creating a forcing function that makes you execute seriously.
"It is only really valuable and things get really real if you burn the boats. That's when I really felt, because we knew that when the company was not working, me and my co-founder were thinking, 'Oh my God, we rejected all these job offers, what are we going to do?' It actually forces you to make things."
However, Varun acknowledges that "burning boats" doesn't mean burning your literal future. "It's not like really, really burning boats, per se, if you have a job, you can get a job. It's not going to go anywhere." The point isn't that failure would be catastrophic—it wouldn't be. Rather, the point is that creating urgency and commitment forces better decision-making and faster execution.
This is particularly applicable in the AI era, where the cost of building has collapsed. "Especially with AI, the cost of building things is so low. People should just build things and try to deliver as much value as they can to a very small set of customers and see if they can pay them money."
The barrier to entry for building and testing new ideas has never been lower. A small team can use AI tools to build prototypes, create initial products, and test market demand with minimal capital expenditure. The only real investment required is time and commitment. Given how low these barriers are, the only remaining excuse for not starting is overthinking.
Conclusion
The story of how two IIT engineers turned down $550,000 job offers to build Giga represents more than just a startup success story. It's an illustration of how exceptional founders think differently about risk, value creation, and what's worth their limited time on Earth.
When Varun received that offer from a top quantitative trading firm and his co-founder received similarly impressive opportunities, they faced the same choice that many talented engineers face: take the known path to financial security or take a chance on something uncertain. Their choice to build Giga, shaped by Y Combinator's mentorship and their own ambition, created a company that's now transforming how enterprises deploy AI.
The key insights from their journey—that product matters more than sales in AI markets, that early customer validation beats endless refinement, that building close to your customers matters, that hiring exceptional people matters more than hiring many people, and that burning your boats forces you to execute—apply far beyond just Giga. These principles address fundamental questions about how to build successful companies in the AI era.
Today, Giga serves some of the world's largest companies, dramatically improving customer support experiences while reducing costs. This achievement came not from having a perfectly executed plan from the beginning, but from starting despite uncertainty, pivoting when evidence suggested a better direction, and relentlessly focusing on whether customers would actually pay for what they built.
For aspiring founders reading this, the final message is clear: the cost of trying has never been lower, the tools available have never been better, and the potential impact has never been higher. Stop overthinking, start building, find customers willing to pay, and see how far you can push your limits. That's how Giga started, and it's how the next generation of transformative companies will emerge.
Original source: Why Two IIT Engineers Turned Down $550K Jobs To Build A Startup
powered by osmu.app