Espresso AI Renovates Snowflake Warehouses with Kubernetes

Espresso AI, a new platform from ex-Googlers that utilizes large language models (LLMs) for data warehouse cost optimization, today announced the launch of a Kubernetes Scheduler for Snowflake. This innovative tool intelligently directs queries between warehouses in real time, addressing long-standing challenges faced by Snowflake users.

Founded by former Google DeepMind engineers, Espresso AI leverages cutting-edge AI research to enhance utilization and cut costs by up to 50% for Snowflake customers. The company's new Kubernetes Scheduler acts as a proxy, dynamically routing queries based on available computing resources, improving performance, and reducing unnecessary expenses.

Solving a Critical Industry Problem

Snowflake users have traditionally grappled with the absence of dynamic scheduling between warehouses. This often forces organizations to choose between overprovisioned, costly warehouses or fragmented, underutilized resources that hinder performance. Espresso AI's solution bridges this gap by enabling real-time, intelligent workload allocation.

"When users or tools send a query to Snowflake, our AI-driven agent determines the best warehouse to handle it, ensuring optimal resource use," explained Ben Lerner, CEO and Co-Founder of Espresso AI. "Think of it like Uber Pool for your queries—if there's space in an existing warehouse, we make sure your workload uses it, maximizing your current compute capacity."

The core innovation lies in decoupling logical compute from physical resources. Instead of dedicating fixed warehouses to specific workloads—which can lead to inefficiency and higher bills—the scheduler maps each query to any warehouse with available capacity. When no suitable warehouse exists, the system automatically spins up a new one and subsequently shuts it down once the workload completes, resulting in significant cost savings.

Real-World Impact

Customers are already experiencing benefits. Tim Hsu, Engineering Manager at Goldbelly, shared his experience, saying, "Espresso AI was incredibly easy to integrate and delivered immediate, measurable savings on our Snowflake spend. It's one of the highest ROI initiatives we've implemented."

Espresso AI's founders—Ben Lerner, Alex Kouzemtchenko, and Juri Ganitkevitch—bring extensive experience from Google Search, Google Cloud, and Google DeepMind's research on machine learning, systems performance, and deep learning. The company has secured $11 million in seed funding from early investors, including FirstMark Capital, Nat Friedman, and Daniel Gross.

It's not every day that a startup offers such a clear value proposition. Espresso AI can be flipped on and off so that teams can see for themselves whether savings are happening. This could easily be hundreds of thousands of dollars in savings for scaling companies—sometimes even millions of dollars. If you're a current Snowflake user, it seems like Espresso AI is a no-brainer.