Today, we are dropping another episode in our “chats” series, specifically on the founder side – hearing from those scaling the companies themselves.

In this episode, we are talking with Daulet Amirkhanov, Founding Engineer of Bead AI. Daulet is going to take us through his years at Meta and Cognee, leading into how he is building Bead AI, to take on compliance audits and AI automation.

Questions

• Tell me and my audience a little bit about you. You’ve gone from three years on high-throughput reliability infrastructure at Meta, to engineering the GraphRAG engine and semantic memory systems at Cognee, and you’re now Founding Engineer at Bead AI — an a16z-backed startup building autonomous agent infrastructure for compliance audits. How did that journey shape the way you think about engineering for the age of autonomous systems?
• Let’s zoom into the Meta years. For listeners who haven’t worked at that scale — what was the exact piece of logging and reliability infrastructure you owned, what does “high-throughput” actually mean in numbers there, and what’s one specific architectural decision from those years that still shapes how you build today?
• A lot of infra engineers stay in infra. You made a deliberate move from human-scale systems at Meta to agent-scale systems at Cognee. What did you see in that moment that convinced you AI agent infrastructure was the next distributed systems frontier — and not just the current hype cycle?
• Cognee is a GraphRAG and semantic memory company, and your work there was on the agent infrastructure side. Your biggest design call was decoupling the MCP architecture so multiple agentic systems can share unified memory through a standalone process, rather than each one coupling to its own Python runtime. Walk us through what problem that was solving and the key design decision you made.
• Give us a concrete example: an agent task that breaks when each agent has its own vector store, but works once they share unified state through the decoupled MCP architecture you built. What’s the actual mechanism that makes the difference?
• Most engineers in this space come from an ML or applications background. You’re coming at agent infrastructure from a pure distributed systems lens. What does that lens let you see that the ML-native crowd is missing?
• Bead is a16z-backed and going after compliance audits, which isn’t the obvious first market for autonomous agents. You joined as Founding Engineer in January and are shaping the technical core now. From your seat: what makes compliance audits the right wedge for agent infrastructure, and what are the foundational decisions you’re making today that will define what the product can do two years from now?
• Make a technical claim about agent infrastructure that most people in this space would push back on — and defend it. Where are you the dissenting voice?
• Without breaking anything confidential — what’s the hardest unsolved problem on your plate at Bead AI right now, and how are you approaching it?
• Two years from now, what’s the piece of agent infrastructure that we’ll consider “obviously necessary” but doesn’t exist yet? Who builds it, and what does it look like?

Sponsors

• Unblocked
• Braingrid.ai
• .TECH Domains
• Mezmo

Links

• https://usebead.ai/
• https://www.linkedin.com/in/amirdnur/