AI is Hard Because Data is Hard

If Humans Can't Understand Enterprise Data, Neither Can AI

Cocoon Catalogs Data for AI

Thorough Enough to Answer >95% of Ad Hoc Data Questions

Get Started

Open Source

Full Features

Cocoon is open-source. Try out Cocoon in Google Colab.

Cocoon turns your DBT projects to data catalog, for GenAI (e.g., GPT-4, Claude-3, Gemini-Ultra, or your local LLMs).

Visit GitHub Google Colab

Need support or have questions? Contact Us

Examples of How Cocoon Boosts AI

[Amazon Ads] How many active keywords are associated with each campaign? [Amazon Ads] What is the average revenue per user for each ad campaign? [Amazon Ads] What's the active target keyword that drives the most clicks? [Amazon Ads] What's the name of the account that has the most active ads? [Amazon Ads] Which ad groups have the highest number of targeted keywords? [StreetEasy]How did the Prospect Park area's price index change MoM in October 2017? [StreetEasy] In which year did the Bronx experience the highest growth in its real estate index? [StreetEasy] What are the price indices for housing in NYC boroughs? [StreetEasy] Which borough had the highest total inventory in June 2018? [StreetEasy] Which neighborhoods in Brooklyn had the highest year-over-year change in rental prices as of September 2018? [StreetEasy] Which quintile of properties in Queens showed the highest MoM price index change in September 2018?

Research

GenAI is based on the following research papers:

Data Ambiguity Strikes Back: How Documentation Improves GPT's Text-to-SQL

@inproceedings{huang2023data,
    title={Data Ambiguity Strikes Back: How Documentation Improves GPT's Text-to-SQL},
    author={Huang, Zezhou and Damalapati, Pavan Kalyan and Wu, Eugene},
    booktitle={NeurIPS 2023 Second Table Representation Learning Workshop},
    year={2023}
}