Mirendil Raises $200M Seed Round to Build the AI That Builds AI

Mirendil has entered the frontier AI conversation with one of the largest seed financings the industry has seen. The San Francisco-based company announced a $200M seed round led by Andreessen Horowitz (a16z) and Kleiner Perkins, with NVIDIA participating. The round values Mirendil at $1B, signaling extraordinary investor confidence in a company that has yet to disclose commercial customers or revenue.

The company was founded in 2026 by Behnam Neyshabur (Co-Founder & CEO), Harsh Mehta (Co-Founder & CTO), Shayan Salehian (Co-Founder), and Tara Rezaei (Co-Founder). Collectively, the founding team brings experience from Anthropic, Google DeepMind, xAI, and OpenAI, with a stated mission that reaches well beyond launching another AI application. Mirendil wants to build systems that accelerate AI research itself.

That distinction matters. The market is crowded with companies building on top of foundation models. Mirendil is trying to influence how the next generation of those models gets built.

What Happened

Mirendil announced a $200M seed financing led by Andreessen Horowitz (a16z) and Kleiner Perkins, with NVIDIA joining as a strategic investor. The company is building frontier AI systems designed to perform AI research and development by automating increasingly large portions of the research process rather than simply generating outputs for end users.

Headquartered in San Francisco, Mirendil has assembled a founding team of approximately 20 researchers and engineers from organizations including Anthropic, Google DeepMind, xAI, and OpenAI. Rather than positioning itself as another model provider, the company is focused on creating systems capable of iterating through the AI development lifecycle itself. Its mission is to democratize frontier AI research by making advanced research capabilities accessible beyond a handful of elite laboratories.

That ambition immediately separates Mirendil from much of today's AI startup landscape. Plenty of companies are competing to build better assistants. Far fewer are trying to build better scientists.

Why This Matters

Money often follows momentum. This round suggests investors believe something different is happening.

Mirendil has not disclosed revenue, customer counts, or commercial deployments. Instead, investors are backing technical credibility, founder experience, and a clearly defined thesis.

Modern frontier AI research requires extraordinary infrastructure, specialized talent, and enormous computational resources. Those ingredients remain concentrated inside a small number of organizations. Mirendil believes much of that research process can itself become increasingly autonomous.

If successful, the company would not simply make researchers more productive. It could fundamentally reduce the barriers preventing organizations outside the largest AI labs from conducting frontier research. That is a much larger opportunity than another workflow assistant.

Market Context

The AI market has entered a new phase. The first wave focused on foundation models. The second wave centered on applications. A third wave is emerging around infrastructure that enables the creation of better models themselves.

Mirendil fits squarely inside that category. Its vision extends across proposing research experiments, writing and executing code, evaluating results, managing compute resources, comparing checkpoints, and determining subsequent research directions with increasing autonomy.

Think of today's AI development process as a laboratory filled with highly skilled researchers moving between whiteboards, servers, datasets, experiments, and evaluation frameworks. Mirendil is attempting to build systems capable of participating across that entire workflow. That is a significantly harder problem than generating text or answering questions. It is also one that could create enormous leverage if solved.

Competitive Landscape

The founders' backgrounds explain much of the excitement surrounding the company. Behnam Neyshabur previously worked at Anthropic and Google DeepMind and is widely recognized for influential contributions to machine learning research. Harsh Mehta helped build Anthropic's internal autoresearch platform before co-founding Mirendil. Shayan Salehian brings machine learning engineering experience from xAI, while Tara Rezaei contributes an academic perspective that Andreessen Horowitz highlighted during its investment announcement.

Collectively, that experience spans several of the organizations shaping modern frontier AI.

Equally notable is what Mirendil has chosen not to emphasize. There are no splashy consumer product announcements. No viral chatbot demonstrations. No inflated customer metrics. Instead, the company is asking investors to back a difficult technical problem whose commercial impact could compound for years.

Sometimes the loudest signal in venture capital is what founders deliberately avoid talking about.

What This Signals

The financing reflects a broader shift across venture capital. Elite investors are increasingly willing to fund exceptionally technical founding teams before traditional business metrics emerge when those teams are addressing foundational infrastructure problems.

Andreessen Horowitz, Kleiner Perkins, and NVIDIA are investing in a hypothesis: that the future constraint in artificial intelligence will not simply be model scale. It will be the speed, quality, and repeatability of AI research itself.

If that proves true, companies building research infrastructure could become just as strategically important as companies building frontier models.

For founders, the lesson is equally important. Technical depth still matters. Clear market positioning still matters. Building something genuinely difficult still attracts capital when the vision is coherent enough to justify it.

The Bigger Industry Shift

Artificial intelligence has always carried a contradiction. Every breakthrough promises greater accessibility, yet the resources required to produce those breakthroughs become increasingly concentrated.

Mirendil is attempting to change that dynamic by building systems that could eventually allow engineers, researchers, scientists, and domain experts across medicine, biology, chemistry, robotics, and other disciplines to access capabilities traditionally reserved for the world's largest AI laboratories.

That makes this financing about more than one company. It reflects growing investor conviction that the next generation of AI value will not come only from better models, but from better systems for creating those models. The conversation is shifting again, from models, to applications, to the infrastructure that builds the systems behind them.