Are soaring AI compute costs slowing down your startup scaling efforts? Join us as we sit down with Junchen Jiang, CEO of TensorMesh, to reveal how his groundbreaking KV cache optimization can slash your AI inference expenses by up to 80 percent.

In this highly technical and insightful episode of Born in Silicon Valley, we explore the rapidly evolving landscape of self-hosted large language models. Junchen explains the mechanics behind KV cache, why traditional GPU memory limitations bottleneck AI performance, and how offloading this data can dramatically improve efficiency for enterprise applications, coding agents, and chatbots.

We also dive into Junchen’s personal journey transitioning from leading a PhD research group at the University of Chicago to building a venture-backed startup in the Bay Area. He shares valuable leadership lessons, the differences between motivating academic researchers versus industry engineers, and what it takes to build a world-class team in today's competitive tech ecosystem.

Whether you are a tech founder, a forward-deployed engineer, or an enterprise leader looking to optimize your AI infrastructure, this conversation is packed with actionable insights on building the next big data layer for artificial intelligence.

Chapters

00:00 Introduction to Junchen Jiang and TensorMesh 02:00 The origin story of KV cache technology 10:07 How KV cache reduces inference costs by 10x 14:54 Building industry-ready systems from research 19:59 Open source community and industry adoption 25:11 Business model and monetization strategies 29:47 Team building and leadership lessons 40:06 Market opportunities and customer segments 44:55 Future vision for AI data layers

Host: Jake Aaron Villarreal leads the top AI recruitment firm in Silicon Valley, www.matchrelevant.com, uncovering stories of funded startups and going behind the scenes to tell their founders’ journeys. If you are growing an AI startup or have a great story to tell, email us at: jake.villarreal@matchrelevant.com