Founder Story: Vipul Ved Prakash of Together AI

The Boy Who Got Goosebumps From Code

Vipul Ved Prakash was born in 1977 in New Delhi, India, into a generation that watched personal computing arrive from a distance. He wrote his first program in BASIC at the age of twelve, and the memory never left him. Decades later, describing that moment to a reporter, he said simply, "It still gives me goosebumps." That early jolt of wonder — the feeling that a few lines of instruction could make a machine come alive — became the through-line of a career built almost entirely on curiosity rather than credentials.

He enrolled at St. Stephen's College in Delhi, studying mathematics, physics, and computer science, and by some accounts also played competitive table tennis. But the classroom could not hold him. Prakash dropped out to chase software design on his own terms, teaching himself the craft in the late 1990s as the internet was exploding into public consciousness. It is a familiar founder archetype now, but at the time, walking away from an elite Indian college to write code independently was a genuine leap of faith. That decision established the pattern that would define him: he learned by building, and he built in the open.

The Anti-Spam Years and the Rise of "Vipul's Razor"

Prakash's first taste of real influence came from an unglamorous problem: email spam. In the early 2000s he created Vipul's Razor, an open-source, collaborative spam-filtering system. The idea was elegantly communal — rather than every server fighting spam alone, users who received junk mail could report it, and that collective intelligence would inoculate everyone else. As he later described it, "I ended up building an open-source collective intelligence system that allowed people who were receiving spam emails to report them." The tool spread virally through the open-source world, eventually running on more than 10 million servers.

The recognition followed. In 2003, at age 25, MIT Technology Review named him to its TR35 list of the top innovators in the world under 35 for his anti-spam work. He turned that reputation into a company: in 2001 he co-founded Cloudmark alongside Jordan Ritter, a co-founder of Napster, to build a commercial messaging-security business on top of the ideas behind Vipul's Razor.

• Vipul's Razor: open-source, deployed on 10M+ servers, built on crowd-sourced reporting.
• Cloudmark (2001): the commercial arm, co-founded with Napster's Jordan Ritter.
• Cloudmark was later acquired by Proofpoint in 2017 for a reported $110 million.

The Cloudmark chapter mattered less for its exit than for what it taught him: that an open, decentralized system, powered by many participants, could outperform closed, centralized ones. That lesson would echo through everything he built next.

Topsy, and the Road to Apple

In 2007, Prakash co-founded Topsy, a social-media search and analytics company, right as Twitter and the real-time web were reshaping how information moved. Topsy did something technically audacious for its era: it built an indexing system from scratch capable of ingesting the firehose of social media at massive scale. "We built an indexing system from the ground up that indexed over a trillion social posts and was up-to-the-second real-time," Prakash recalled. In a world where search meant crawling static web pages, indexing a trillion posts in real time was a systems-engineering feat.

That capability caught the attention of Apple. In December 2013, Apple acquired Topsy for a reported $200 million-plus, one of Apple's larger publicly reported acquisitions of the period. Prakash did not simply cash out and disappear.

Apple did not just buy a product; it bought a systems architect who understood search, indexing, and machine learning at planetary scale.

He stayed at the company for roughly five years, until 2018, serving as a director of engineering. There he helped build search technology woven into products like Spotlight and Siri, and he appeared on stage at WWDC 2015 to introduce Apple's iOS Search APIs. Working inside one of the most closed, vertically integrated companies on earth gave him an intimate view of how much power accrues to whoever controls the full stack — a view that would shape his eventual discomfort with a closed AI future.

Founding Together AI

By 2022, generative AI was cresting, and Prakash saw both the promise and the danger clearly. Foundation models represented a generational shift in technology, but the economics of training them — enormous clusters of expensive GPUs — were concentrating that power in the hands of a few well-capitalized labs. In June 2022, he co-founded Together AI to build a decentralized, open alternative.

He did not do it alone. Together AI's founding team reads like a who's-who of AI systems research:

1. Vipul Ved Prakash — Founder & CEO, the serial entrepreneur and systems builder.
2. Ce Zhang — Founder & CTO, a machine-learning-systems researcher (Stanford post-doc, later ETH Zürich and University of Chicago).
3. Percy Liang — Founder, the Stanford professor behind foundational work on foundation models and evaluation.
4. Chris Ré — Founder, the Stanford professor whose prior startup Lattice.io was acquired by Apple in 2017.
5. Tri Dao — Founder & Chief Scientist, the researcher who created FlashAttention, a breakthrough that made training large models dramatically more efficient.

The diagnosis at the heart of the company was infrastructural. "The cloud was built for CPU computing. GPU access was expensive and limited," Prakash observed. Together AI set out to build a full-stack AI cloud — renting and orchestrating Nvidia GPU clusters and layering its own systems research on top — so that any developer could train, fine-tune, and run open-source models at high performance and low cost. Its mission, in the company's own words, is a "full-stack AI platform powered by cutting edge system research," aimed squarely at democratizing access to the compute that had become the scarcest resource in AI.

The Open-Source Thesis

If there is a single conviction that animates Prakash, it is the belief that the most consequential technology of our era must not be controlled by a handful of gatekeepers. He has stated the stakes bluntly:

"This is the most important technology humans will build. And it's a huge risk if the best models are closed and controlled by just one or two companies."

Together AI put that thesis into practice by contributing to the open ecosystem, not just profiting from it. In 2023 the company released RedPajama, an open reproduction of a large-scale pretraining dataset, followed by RedPajama-V2, a corpus of over 30 trillion tokens that was downloaded more than a million times in a single month. The message was unmistakable: Together did not merely want to sell GPU time, it wanted to arm the open-source community with the raw materials to compete with closed labs.

Prakash frames the ambition as building the home base for open AI development. "We want to be the platform that provides access to tier-one, open-source models," he has said, and, more expansively, "We want to be the place where developers build and do research and then deploy their AI systems. We think it's a massive market." It is a strategy that treats openness not as charity but as a competitive advantage — a bet that a decentralized, transparent ecosystem will out-innovate closed silos over the long run, just as Vipul's Razor once out-filtered proprietary spam tools two decades earlier.

Defining Moments: The Numbers Behind the Bet

Ideals are easy; capital markets are the harsh test. On that measure, Together AI's rise has been extraordinary. The funding trajectory tells the story of a thesis rapidly winning converts:

• November 2023 — Series A: $102.5 million led by Kleiner Perkins, with Nvidia and Emergence Capital, at roughly a $500 million valuation.
• March 2024 — Series A extension: an additional $106 million led by Salesforce Ventures, pushing the valuation to about $1.25 billion.
• Early 2025 — Series B: $305 million led by General Catalyst, at a $3.3 billion valuation.
• July 1, 2026 — Series C: $800 million at an $8.3 billion valuation, led by Aramco Ventures, with Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, March Capital, Pegatron, and SentinelOne's S Ventures participating.

The Series C more than doubled the company's valuation in roughly 16 months. Just as striking are the business fundamentals underneath it: annual bookings crossed $1.15 billion in the most recent quarter, up from an annualized run rate near $100 million only about a year and a half earlier. The customer roster — reportedly including fast-growing AI-native companies like Cursor, Cognition, and Decagon, alongside enterprises such as Zoom and Quora — signals that open-source infrastructure has moved from ideological preference to production necessity. The macro tailwind Prakash bet on has materialized: by the company's account, open-source model usage across the industry roughly tripled in twelve months.

Industry Impact and Legacy

Prakash's larger significance is that he has helped make "open-source AI" a viable commercial category rather than a hobbyist ideal. In a market where the loudest narrative has been that only a few trillion-dollar companies can afford to build frontier models, Together AI's ascent is a working counter-argument. By packaging GPU access, efficient training and inference, and open datasets into a single platform, it has lowered the barrier for hundreds of thousands of developers and enterprises to build on models they can inspect, modify, and own.

He also embodies a distinctly recognizable arc among Indian-origin technologists now shaping the global AI industry — a self-taught outsider who repeatedly built open, decentralized systems and turned them into companies acquirers and investors chased. His career forms a clean line: Vipul's Razor's crowd-sourced spam defense, Topsy's real-time index of a trillion posts, and now an open cloud for foundation models. Each was a systems-scale bet that distributed, transparent architectures beat closed ones.

From spam filters to social search to foundation models, Prakash has spent a quarter-century building the same idea in different clothes: that keeping powerful technology open is not only more principled — it is more powerful.

Closing Thoughts

There is a tidy symmetry to Vipul Ved Prakash's story. The twelve-year-old who got goosebumps from a BASIC program grew into a founder still chasing that same feeling at the largest possible scale — and still insisting the tools be shared. He did it without the safety net of a finished degree, learning in public, shipping open-source code, and letting the work earn his reputation. The MIT innovator award, the Apple acquisitions, and now a multi-billion-dollar AI cloud are the byproducts, not the point.

What makes his current chapter compelling is the size of the wager. Prakash is not merely building a profitable infrastructure business; he is trying to prove that the most important technology humanity will build can remain open, decentralized, and broadly accessible — that it need not calcify into the hands of one or two companies. Whether the open ecosystem ultimately rivals the closed frontier labs is one of the defining questions of this decade. But if it does, a large part of the reason will be that a self-taught programmer from Delhi kept betting, over and over, on the same stubborn conviction: that the best systems are the ones everyone can build on together.

Your Story Could Be Next

Every founder in this series started exactly where you are now — with an idea and the will to build. What has changed is the leverage. Frederick gives you a team of AI agents that take you from idea to a real company, helping you design, build, and ship software on demand — no giant team and no endless runway required.

If Vipul Ved Prakash's journey resonated, there is no better time to begin your own. Start building your company with Frederick.

References

1. https://en.wikipedia.org/wiki/Vipul_Ved_Prakash
2. https://techcrunch.com/2026/07/01/neocloud-together-ai-raises-800m-leaps-to-8-3b-valuation/
3. https://www.businesswire.com/news/home/20260701243402/en/Together-AI-Raises-$800-Million-at-$8.3-Billion-Valuation-to-Make-Frontier-AI-Accessible-to-All
4. https://www.forbesindia.com/article/ai-special-2025/togetherais-vipul-ved-prakash-democratising-ai-access-with-opensource-solutions/96265/1
5. https://research.contrary.com/company/together-ai
6. https://www.together.ai/about-us