Mirendil raises $1 billion to automate AI model development
Mirendil Inc. has raised $1 billion in a Series A funding round led by Kleiner Perkins and Clearlake Capital, with Google as a key investor. The startup’s Lor platform automates AI model training using hundreds of AI agents, enabling a single researcher to build a competitive coding model in 3.5 months for $10 million.
Editor, Lazyfounder
TABLE OF CONTENTS
- $1 billion Series A round led by Kleiner Perkins and Clearlake Capital
- Founders and platform designed to industrialize AI development
- Automation cuts development time and cost by an order of magnitude
- Scaling plans and strategic implications for AI labs
- How Mirendil’s technology fits into the broader AI automation trend
- What this means
- Key takeaways
- Sources
$1 billion Series A round led by Kleiner Perkins and Clearlake Capital
Mirendil Inc., a startup focused on accelerating artificial intelligence development through automation, has raised $1 billion in a Series A funding round. The round was jointly led by Kleiner Perkins and Clearlake Capital, with participation from Google and more than half a dozen other investors.
Founders and platform designed to industrialize AI development
The company was founded last year by Behnam Neyshabur, now CEO, and Harsh Mehta, CTO. Both are former researchers at Google DeepMind and Anthropic PBC. Neyshabur co-invented the SAM algorithm, which improves AI model performance on untrained tasks, while Mehta specialized in pretraining projects at his previous roles.
Mirendil’s first product, Lor, is a platform that automates AI model training by distributing tasks across hundreds of AI agents. The platform handles trial-and-error cycles, error correction, and optimization, allowing researchers to focus on higher-level decisions.
Automation cuts development time and cost by an order of magnitude
Mirendil demonstrated Lor’s capabilities earlier this year by tasking a single researcher with building a coding model to rival leading open-source large language models. The project, completed in 3.5 months at a cost of $10 million, relied on Lor’s automation for over 99% of the code, infrastructure setup, algorithm selection, and training data collection.
The resulting model, Mirendil-Base, serves as the foundation for Fern, a specialized coding model capable of processing 10 million input tokens per prompt—ten times the capacity of Claude Opus 5.5. Fern is designed to address common programming challenges faced by AI labs.
Scaling plans and strategic implications for AI labs
The funding will be used to scale Lor and develop more advanced large language models. Late last month, Mirendil previewed upgraded versions of Mirendil-Base with parameter counts ranging from 500 billion to 10 trillion, signaling its ambition to compete in the high-end model space.
Google’s involvement as an investor highlights the growing alignment between AI development platforms and cloud providers. This may accelerate integration with hyperscale cloud services, offering startups and enterprises faster, cheaper pathways to custom AI models.
How Mirendil’s technology fits into the broader AI automation trend
Earlier Lazyfounder coverage showed Google Cloud embedding autonomous AI agents into its enterprise tools, reducing the need for manual oversight in long-running tasks. Mirendil’s Lor platform takes a similar approach but applies it to the AI development lifecycle itself, automating tasks that traditionally require large teams of researchers.
Google DeepMind’s OpenRL, an open-source tool for reinforcement learning on Kubernetes, complements Lor by simplifying the infrastructure layer. Labs using both could reduce operational complexity while accelerating model iteration.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
Mirendil’s $1 billion Series A is a signal for AI infrastructure startups. The round—led by Kleiner Perkins and Clearlake Capital, with Google as a backer—validates the idea that AI model development can be industrialized. Founders building workflow tools for AI teams now have a benchmark: automation platforms that cut both time and labor costs by an order of magnitude (one researcher, 3.5 months, $10 million) are fundable at unicorn scale.
Key takeaways
- The $1 billion Series A sets a new bar for AI infrastructure startups, showing that investors will back platforms that automate core AI development workflows.
- Mirendil’s claim—that one researcher built a coding model in 3.5 months for $10 million—challenges the conventional wisdom that large teams and nine-figure budgets are prerequisites for competitive LLMs.
- Google’s participation underscores the strategic overlap between cloud providers and AI development tools; operators should expect tighter integration between AI training platforms and hyperscale cloud services.
- Fern’s ability to process 10 million input tokens per prompt makes it a direct competitor to closed models like Claude Opus 5.5, offering a 10x context advantage for labs that need long-context coding assistance.
Related on Lazyfounder
Sources
- SiliconANGLE · 2026-11-13
Mirendil raises $1B in funding to speed up AI model development
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.
