Aravind
Srinivas

The Answer Engineer

Somewhere around 2022, a new Perplexity employee asked a simple health insurance question. Google returned ads. The AI Slack bot hallucinated. Aravind Srinivas watched both happen and thought: citations. If academic papers can back every claim with a source, why can't AI?

That thought - borrowed from the rigor of research papers he had spent years reading and writing - became the founding principle of Perplexity AI. Every answer gets citations. Every claim is traceable. The internet's oldest problem - you don't know if what you're reading is true - gets a structural fix, not a hope-for-the-best one.

Perplexity launched publicly on December 7, 2022. Seven days after ChatGPT. In the AI gold rush of that winter, launching a week after the most-discussed product in the history of the internet might look like terrible timing. It turned out to be perfect. Srinivas wasn't competing with ChatGPT. He was solving a different problem - not "generate text" but "find truth and show your work."

The 0.01 That Changed Everything

In 2017, Aravind Srinivas was finishing his dual degree in Electrical Engineering at IIT Madras. He wanted to switch to Computer Science. The cutoff required a certain CGPA. He missed it by 0.01.

Not 1. Not 0.1. Point zero one.

Most people would have been crushed. Srinivas took it as a redirect. Without access to the official CS program, he self-taught deep learning. He read papers obsessively. He developed a researcher's instinct, not a student's one. When he arrived at UC Berkeley for his PhD - working under Pieter Abbeel, one of robotics and ML's most respected minds - he already had the intellectual habits of someone who had been learning outside the system. His thesis on representation learning for perception and control drew from three major AI labs. He wasn't just reading about the field - he was building it.

The cruel math of that 0.01 CGPA is, in retrospect, one of the stranger origin stories in Silicon Valley. The door that didn't open pushed him toward a window that looked out onto something much larger.

The Lab Collector

Between 2018 and 2021, Srinivas did something almost no one had done: he worked, in sequence, at all three of the world's most significant AI research labs. OpenAI in 2018, where John Schulman - the researcher who would later be instrumental in creating ChatGPT - noticed his work on policy gradient algorithms. DeepMind in 2019, contributing to CPCv2, a foundational contrastive learning paper. Google Brain in 2020 and 2021, where he worked alongside Ashish Vaswani, the architect who helped invent the transformer - the mechanism now sitting inside virtually every major AI product on earth.

He then went full-time at OpenAI in 2021, working on language and diffusion models. He was inside the rooms where modern AI was being assembled.

Most engineers who pass through those labs stay. The access, the compute, the colleagues - it's a gravitational field. Srinivas left. He describes himself as "unemployable" - not as a boast, but as a diagnosis. He prefers answering to users over answering to bosses. That preference, it turned out, was the company-shaped hole he needed to fill.

Building Without a Deck

Perplexity's founding round came together in 2022. Jeff Bezos eventually wrote a check. Nvidia followed. SoftBank Vision Fund. IVP. Accel. By early 2026, total funding had exceeded $1 billion across eight rounds, and the valuation had climbed from $150 million at seed to $21.21 billion.

Srinivas has never made a pitch deck for any of them. He writes memos instead - dense, direct, no slide template forcing ideas into three bullet points. "Write a memo," he tells people at Berkeley Haas, "and ask whatever you want." The discipline of memo-writing, he thinks, forces clarity that slides actively prevent. When your argument has to survive as a document rather than a presentation, weak parts become obvious.

That approach extends to how he thinks about products. Perplexity's subscription model - not an ad-supported one - is a structural decision, not a financial preference. Ads align incentives with clicks, not with correct answers. Subscriptions align incentives with accuracy. If the product is wrong too often, subscribers leave. If the product is right, they stay. The feedback loop is clean.

He is relentless about speed. He ships at 80% quality and iterates. He tests Perplexity personally, often on airplane WiFi - the worst possible internet conditions - to ensure the product holds up where it matters least forgiving. Larry Page used to test Chrome on ancient laptops. Srinivas has internalized the same principle: optimize for constraint, not for optimal conditions.

Taking Up Space

In August 2025, Perplexity submitted an unsolicited $34.5 billion cash bid to buy Google Chrome. At the time, Perplexity itself was valued at less than that. The bid came amid antitrust proceedings against Google over its browser dominance. Srinivas pledged to keep Chromium open source, invest $3 billion over two years, and not change Chrome's default search engine. The move drew worldwide coverage, most of it incredulous.

That is, broadly, Srinivas's competitive posture. He doesn't position Perplexity as a slightly better Google. He asks different questions - why do search results have to be a list of blue links? Why does the prominent real estate go to links rather than answers? Why can't the journey begin after you get the answer, not before?

In January 2025, Perplexity submitted a merger proposal for TikTok, the day before the US ban was set to take effect. The revised offer included giving the US government up to 50% ownership upon IPO at a $300 billion-plus valuation. It didn't go through - but Srinivas had made the point that Perplexity was thinking about information infrastructure, not just query volume.

The Curiosity Engine

There is a phrase Srinivas returns to across interviews, talks, and posts: "serve the world's curiosity." It is Perplexity's stated mission, and he treats it as a genuine philosophical commitment rather than a marketing line. He quotes it at Stanford GSB. He grounds it in his own history - the student in Chennai who read obsessively, the PhD student who spent an internship in Bengaluru almost entirely indoors, nose in research papers while the city moved outside his window.

"Curiosity is the number one condition for success," he posted on X. "Anyone who makes anything important is curious. If you're not curious, you're an NPC."

His wife, Sharmada Vishwanath, is an Indian classical dance professional. They married in November 2021, just as Perplexity's founding was taking shape. He credits her with keeping him grounded - someone whose work exists in a completely different register from AI research, who brings him back from abstraction to presence. He was raised vegetarian in Chennai, now practices ovo-vegetarianism, comes from a middle-class family where his mother's encouragement to attend IIT was the original push toward the path that led here.

The Hurun India Rich List placed his net worth at $2.5 billion in October 2025. India's youngest billionaire. He was 31. The figure reflects his equity stake in Perplexity. He has been angel investor in ElevenLabs, Pika Labs, Suno, Cursor, Mistral AI, and others - the kind of early-stage bets that a researcher with an eye for technical originality makes when they have conviction about where things are going.

What He's Actually Building

Perplexity in early 2026 is not just an answer engine. It is Comet, a browser with AI built in from the ground up. It is Perplexity Computer, an AI agent that handles end-to-end projects. It is Perplexity Pages, where collaborative AI-generated articles live alongside citations. It is a Publishers Program that shares revenue with the content creators whose work trains the system. It is a partnership with Bharti Airtel, reaching 360 million users in India.

Srinivas's vision for AI moves beyond the current generation of models. He talks about AI that could "disappear for a week like Einstein" - given enough compute and time to reason - and return with a genuinely novel answer. Not a fast retrieval. Not a confident-sounding summary. A breakthrough. That requires inference compute at a scale that doesn't exist yet. He is building toward it anyway.

The number Perplexity reports is 780 million queries per month and growing at 20% per month, processed by roughly 247 people. The ratio - that headcount to that query volume to that valuation - is the operational expression of what Srinivas means when he says "a better product should be one that allows you to be more lazy, not less." He means the machine should do the work. He means every design decision should be in service of the user getting to the answer faster, with more confidence, with less friction.

The 0.01 CGPA that didn't let him officially study computer science. The internship where he barely left the building. The health insurance question that nobody could answer correctly. The citation principle lifted from academic convention and dropped into a consumer product. Everything converges on the same obsession: getting to the truth, and showing that you did.