When Karan Sirdesai talks about AI, he’s not talking about another chatbot or copywriting tool. He’s building the infrastructure that could power the next generation of AI-native applications and he’s doing it through Mira Network.
In our conversation with Karan on the Beacon Podcast, he shared his founder journey, the vision behind Mira, and how decentralization plays a crucial role in scaling AI. Here’s how a curious kid from India grew into a founder reshaping how machine intelligence is built and delivered.
From Stanford to Startups: The Genesis of a Builder
Karan’s interest in systems and infrastructure began at Stanford, where he worked closely with professors and researchers pushing the limits of computer science. But it was his work at Metamind (later acquired by Salesforce) and time spent in the Valley’s startup trenches that crystallized his passion: building infrastructure for intelligent systems.
He noticed a pattern. “Each time I saw a new AI system, it was limited by either the lack of good data, the lack of good infrastructure, or both,” he shared. This realization became the seed for Mira Network.
“I was lucky to work with some amazing professors at Stanford, and I was exposed to both the theoretical and the practical sides of AI very early,” Karan said. “But what really stood out to me was how quickly the systems broke when you tried to scale them.”
What is Mira?
Mira Network is a decentralized infrastructure platform designed to make it easier to build AI-native applications. Think of it as the AI equivalent of AWS, but open, permissionless, and composable.
At its core, Mira provides:
- A library of modular workflows (called AI pipelines)
- Access to pre-configured tooling
- An execution environment for running AI tasks on-chain
- A marketplace for decentralized AI components
The idea is to lower the barrier to entry for developers who want to build AI-powered apps but don’t want to reinvent the wheel for every use case.
“Right now, building a good AI application is like assembling IKEA furniture blindfolded,” Karan joked. “We want to give developers the right toolkit so they can go from zero to production quickly.”
Building in Public: Klok and Early Experiments
Mira’s first ecosystem product is Klok, a crypto-native chatbot built using over 500 datasets. It’s a front-end experience that showcases what’s possible with Mira’s backend. The idea? Test, experiment, and gather feedback.
“Klok is kind of our version of a showcase car,” Karan explained. “It’s not meant to be the final product, but it shows what you can do with our pipelines.” Users can query Klok across multiple blockchains and ecosystems, ask protocol-specific questions, and even get developer support. It’s AI with actual utility, not just a novelty.
“We trained Klok on some of the most active protocol documentation, Discord threads, and developer support chats,” Karan said. “So it’s actually helpful, it’s not just another cute chatbot.”
Decentralized Intelligence: Why AI Needs Web3
Karan is a firm believer that the future of AI needs to be decentralized.
“We’re currently in a world where 2 or 3 companies control access to the best models, the best data, and the best infrastructure. That’s dangerous,” he warned.
Mira flips the model by making these components available on-chain. That includes AI agents, datasets, and execution pipelines. The long-term goal? An open marketplace where AI developers and model creators can share, improve, and monetize their work, without needing to go through Big Tech.
“We want to create an economic layer for AI,” Karan said. “One where creators of models, curators of data, and engineers of workflows all get to participate in the value they help generate.”
AI Agents and Autonomous Systems
Another key concept Karan explores is the rise of autonomous AI agents. These are bots that can take actions, interact with smart contracts, make decisions, and improve themselves over time.
“The next logical step is agentic AI. Not just a chatbot, but an agent that can monitor a DeFi position, place trades, and rebalance a portfolio,” Karan explained. “And we’re building the rails for that.”
This vision of AI aligns closely with crypto: autonomous, decentralized, and self-improving.
“We’re not building just for GPT integrations, we’re building for agents that can be composable primitives in DeFi, gaming, governance, and beyond,” he added.
Challenges of Building at the Edge
Karan is honest about the challenges. AI is a fast-moving field, and aligning that with crypto’s composable but slower-paced infrastructure isn’t easy.
“Right now, if you’re an AI founder, you either pick performance or decentralization—you can’t have both. We’re trying to change that.”He emphasized the importance of user experience, developer tooling, and real-world applications. “Nobody wants another demo that works once and breaks the moment you try something new,” he said.
“Being pragmatic is underrated. We’re okay with being a little opinionated in our infrastructure choices, as long as it helps developers actually ship,” he noted.
Advice for Builders and Founders
So what does Karan advise for aspiring AI or Web3 founders?
“Pick hard problems. Build for the long term. Don’t chase hype,” he said. He believes that real innovation happens when you solve fundamental problems, not just jump on trends.
He also stressed the importance of shipping early and iterating fast. Mira’s approach -build in public, get feedback, and refine – has been central to their traction.
“You don’t need to have everything figured out,” he shared. “You need to have conviction, a good loop of feedback, and the ability to ship again and again.”
The Road Ahead
Mira Network is currently working with developers, researchers, and AI engineers across ecosystems. As they move toward their network launch, the focus is on reliability, usability, and scaling decentralized AI.
“AI isn’t just a tool. It’s a new form of intelligence. And we need to build the infrastructure to support it : ethically, sustainably, and openly.”
Karan Sirdesai is not just building another startup. He’s laying down the rails for a new paradigm in machine learning, one that is open, composable, and powered by crypto. Mira Network is his bet on that future.
