Built for

UNSTRUCTURED

01

USE SQL ON DOCS, WEBS, IMGS, VIDEOS

Roe SQL supports multi-modal data extraction and classification powered by Gen AI models

02

SEMANTIC search VIA where clause

Build multi-modal semantic search indexes on your data. Use them in SQL WHERE clauses.

03

FINE-TUNE VIA HUMAN FEEDBACK

To chase for the extreme consistency, you can fine-tune a task-specific AI model by simply feeding in human feedback.

SIMPLE

DATA JOURNEY


Today, to classify 100 videos, multiple teams have to work together to build an ETL pipeline, build ML models and workflows, and finally put the classification results labels into a structured data warehouse
‍


However, it could have been way simpler. To classify videos with Roe, just one line of SQL
‍
SELECT CLASSIFY("Classify into GOOD or BAD", VIDEO) FROM VIDEOS;
‍

REIMAGINE THE DATA 

USECASES

Financial Documents

Customer Success

Fraud Detection

Social Listening

Frequently

Asked

Questions

01

Aren't lakehouse solution like Databricks already supporting unstructured data?

Databricks (DBx) is extremely tunable through PySpark. In DBx, to process an image, one should configure a Spark cluster, load the data into RDD, pick a model from ML flow, write a python biz logic and finally register it as a SQL function.

For expert ML engineers building models from scratch, DBx is a good choice.

Roe, on the other hand, focuses on the out-of-box time to value.

Since Roe has built-in the multi-modal data AI capability, a simple SQL like this is sufficient SELECT CLASSIFY("Classify into GOOD or BAD", IMAGE) FROM IMAGES;

02

Aren't conventional data warehouses solutions like BigQuery or Snowflake already support unstructured data?

BigQuery and Snowflake are experts in structured data. They're renowned for simplicity, and loved by the millions of data analyst and data scientists. However, since their SQL engine is heavily optimized for text and numbers, unstructured cannot be easily explored, managed and processed like regular tables.

Roe is more than just yet-another-JDBC, but is a product with rich UI so that data people can manage and explore the unstructured data with native AI capabilities

03

I have data stored in Snowflake tables already, how can I benefit from Roe AI?

You can absolutely benefit from Roe AI wherever your data is stored!

Create powerful Gen-AI engines on Roe AI, and make API calls to them directly via Snowflake external access.

04

Can teams with minimal ML background use your platform?

Yes! That's why we build Roe AI. We believe AI belongs to everyone.
Simplicity is in our gene and we want to remove any overhead from the core business data tasks.

05

How much does it cost?

Industry-standard, separate storage and compute usage based billing, prepaid capacity.

We'd love to discuss your budget and curate a personalized capacity plan based on your demand.

About us

Richard and Jason, two engineering minds met at UC Berkeley, later worked at Meta, LinkedIn, Retool, Snowflake, founded Roe AI in 2023.
Richard and Jason have a vision and a principle in building this new data warehouse.
The vision is to make unstructured data accessible to data teams. In the past 20 years, data analytics are all about structured data - numbers and simple texts. With the emergence of Gen AI, unstructured data can be analyzed without involving sophisticated ML models. Roe AI helps data teams leverage the Gen AI efficiently and turn these into strategic decisions.
Their guiding principle revolves around simplicity, a value deeply ingrained from their time at Snowflake. Richard, drawing from this experience, firmly believes in simplicity as a crucial factor for success. With Roe, they aim to uphold this principle by offering an intuitive user experience for data teams and abstracting away the infrastructure.
Users simply load the data and query it—keeping the process straightforward and accessible for all.