Snowflake Startup Spotlight: ROE AI

Welcome to Snowflake’s Startup Spotlight, where we learn about awesome companies building businesses on Snowflake. In this edition, we talk to Richard Meng, co-founder and CEO of ROE AI, a startup that empowers data teams to extract insights from unstructured, multimodal data — including documents, images and web pages — using familiar SQL queries. By integrating AI agents, ROE AI’s platform simplifies data processing, enabling organizations across industries to automate manual workflows and derive actionable intelligence from data.

What inspires you as a founder?

I’ve been driven to challenge the status quo and tackle problems that haven’t been solved before — especially those at the intersection of data and AI. Working at Snowflake before founding ROE AI also inspired me. I experienced the thrilling pace of AI data innovation firsthand.

As soon as large language models (LLMs) emerged, I knew I could create something that addressed a long-standing challenge in the data world: harnessing unstructured data. People have tried to solve it for decades, but the solutions often fell short. I felt this was the perfect time to try something bold, leveraging my background and experience at Snowflake.

What problem does your company aim to solve? 

Many financial services organizations rely on documents for a wealth of insights. Today these workflows are completely manual. My goal is to enable these clients to build agentic, rigorously evaluated document workflows. Many companies try to solve it with RAG, but they often fall short because accuracy is a mandatory requirement. 

That’s why we’re building ROE AI, to enable these clients who care about accuracy to build, evaluate and productionize agentic document workflows in a few lines of SQL. 

And we’re glad to partner with Snowflake and bring this capability to Snowflake clients natively via Snowflake Marketplace.

What’s the coolest thing you’re doing with data? 

ROE AI solves unstructured data with zero embedding vectors. For clients’ unstructured data problems, we do not rely on the vectors because of their imprecise nature. Instead, we use a lot of LLM calls — and we are able to make this process cheap. 

Another interesting fact is that we dogfood our own product for hiring: we use ROE AI internally to screen the resumes of prospective hires. We’re able to quickly parse thousands of resumes in under a minute, answering targeted questions like: Does the candidate have both enterprise and startup experience? Does the candidate bring substantial data expertise?

What are the key benefits you’ve experienced building on Snowflake?

First, Snowflake has enabled us to strengthen user trust in our app. With Snowflake Native Apps and Snowflake Cortex AI, all ROE AI functionality lives within Snowflake’s security perimeter. This means enterprises can run unstructured data workflows, powered by AI agents, without moving data out of Snowflake — which enhances trust and helps support compliance.  

Second, we’re optimizing scalability. Large-scale LLM operations often require specialized resources. With Cortex AI, we can seamlessly scale GPU and other compute resources to handle high volumes of unstructured data analytics. With Cortex LLMs, we can use cutting-edge commercial models securely. 

And third, we’re increasing efficiency. By leveraging SQL functions, Snowflake staging and other Snowflake-native capabilities, end users can query or transform unstructured data using ROE AI in a self-service fashion — exactly the way they query their structured data.

How has the Snowflake Native App Framework shaped your startup's growth and development strategy?

First, we have accelerated our go-to-market strategy through Snowflake’s network. Large enterprises are our ideal customers, but they can be tough for startups to reach. By joining the Snowflake Native App ecosystem, we can tap into Snowflake’s established enterprise audience and sales channels.  

We’re also benefiting from a fully integrated Snowflake query experience. Rather than having a disjointed, stand-alone approach, we’ve embedded our unstructured data solutions into Snowflake’s data transformations and workflows. This tight integration shortens our sales cycles and reduces complexity for the customer.

Finally, our solution already aligns with Snowflake’s security and procurement frameworks, making it easy for customers to buy ROE AI using their existing Snowflake contracts. This can shrink a six-month procurement process to just a few weeks.

What advice would you give to others considering building their apps on Snowflake?

First, embrace Snowflake’s native features fully. Cortex AI, user-defined functions, Snowpark and Secure Data Sharing can significantly compress your development timeline. Second, align your solution with Snowflake’s data security posture to simplify enterprise adoption. And finally, engage early with Snowflake’s partner and sales teams. They have deep insights into customer pain points, which can help you refine your product-market fit and accelerate your go-to-market strategy.

Final question: What's a lesson you learned the hard way? 

Tackle the hard problems first. Chasing easier wins might seem promising initially, but true product-market fit often emerges when we solve the toughest challenges with courage and persistence.

Learn more about ROE AI and its strategy for solving difficult, data-intensive problems with high precision at getroe.ai or try their app on Snowflake Marketplace.