Aug 5, 2024, San Francisco - Roe AI has successfully completed $3.5 million seed round, led by Google’s AI-focused venture fund, Gradient Ventures, with participation from Ardent Venture Partners, Y Combinator, Orange Collective, key executives from Snowflake.

Roe AI's vision is to make any type of unstructured data from any source immediately accessible and usable for data analysts.

‍

Unstructured data is a huge opportunity for data teams

According to a 2019 Deloitte survey, “64% of organizations reported relying primarily on structured data from internal systems and resources. Only 18% have taken advantage of unstructured data (such as product images or customer audio files) or comments from social media.“

Five years later, unstructured data tooling is still in its 1990s. Just think about all these PDF documents companies have been fighting for 30+ years. 

Today, as AI that can is becoming so powerful to see and hear multimodal data like humans, we see two pivotal opportunities for data teams.

First, rethink the form of data that can be used for analysis. For example, events telemetry was the only quantitative way to analyze how users use the product. Today, Gemini can already understand user session replay videos and discover the friction points.

Second, rethink the scope of responsibility within the organization. With multimodal LLM, data teams can build many new 0 to 1 data products based on public webpages data, enterprise operational data, marketing assets data and many more, to make impact in diverse functions directly.

‍

Pioneering a multimodal data query engine for data analysts

We’re pioneering an AI-powered multimodal data query engine, named Volans, that allows data people to manage, enrich, and query large amounts of images, videos, documents, and web pages with only a few lines of SQL queries. Here is our vision:

1. Raw data ELT, not ETL: With the advent of multimodal AI, we believe it’s essential to analyze raw data directly instead of its transformed, opinionated structured form. This gives data analysts much greater flexibility and sovereignty to answer new business questions on multimodal data as needed.

2. Extend SQL for AI: SQL is a language familiar to data professionals. We’re extending its power to query multimodal data with Generative AI agents and multimodal semantic search. It’s as if data professionals have the combined capabilities of Langchain, Pinecone and BigQuery.

3. Lakehouse architecture: Moving unstructured data between storage infrastructure is costly. We bring compute directly to the data storage layer, eliminating the need for data movement.

‍

Accelerating unstructured data adoption to customers and users

We’ve raised more than $3.5 million in funding led by Google’s AI-focused venture fund, Gradient Ventures, with participation from Ardent Venture Partners, Y Combinator, Orange Collective, key executives from Snowflake, and data leaders like Gu Xie from Group 1001 and Daniel Svonava from Superlinked. Our platform is already being used by data teams within Arc, Revere, Crossmint, and several Fortune 2000 companies.

“Unstructured data contains a tremendous amount of business value, but accessing, leveraging, and extracting it has previously been costly, complex and time-consuming.”

Zach Bratun, General Partner at Gradient Ventures

Data teams can build AI agents on Roe AI and directly work with multimodal data using their most familiar language, SQL.

‍

Roe AI, with its AI-powered SQL queries, is helping various industries unlock new use cases out of their multimodal data:

1. Financial data: manage, extract and retrieve from large corpuses of documents. Questions like “find all the Prospectus from Vanguard after 2023” can be made in one SQL query, without preprocessing.

2. Marketing data: enrich the visual and sound elements from video campaign assets and answer questions like “How do product placement choices affect TikTok ads?” which allows them to drive higher ROI on advertising spend.

3. Trust, safety, and KYC: Enrich the visual elements of web pages and social posts so questions like “What is this merchant selling based on its homepage?” can be answered within seconds.

4. AI Copilot Analytics: analyze multimodal chat conversations between AI and humans. This allows the data team to identify high-priority questions like “What are the topics people mostly asked in yesterday’s conversation data?”

5. Enterprise operational data: quickly analyze sales calls from Gong, Zoom, and Slack messages, which allows data teams to answer “What are the inefficiencies in the engineering tooling” or “What are the common asks from the sales calls?”

‍

About the team

Richard and Jason met at UC Berkeley, bringing over a decade of experience building scalable data infrastructure.

Before Roe AI, Richard was one of the founding engineers of Snowflake’s Generative AI arm. In late 2022, Richard saw Gen AI coming for enterprise data and pioneered the engineering development of Snowflake AI Copilot.

Jason Wang was at Retool and Meta, where he was developing high-performance optimization and graph algorithms that supported thousands of ML models and hundreds of ML teams.

Richard and Jason experienced the pain points of multimodal data analysis and decided to solve this issue together.

‍

We’d love to work with you

We’re just getting started on helping every data and AI team to make sense of unstructured data.

• If you’re interested in learning how our platform can help you scale, visit our website getroe.ai and book a demo here.
• If you’d like to join the team, we’re hiring! Find our latest open roles here: https://www.linkedin.com/company/roe-ai/jobs 

‍