Fine-tuned LLMs for enterprise data analysis

Defog lets your business users query data in seconds, using everyday language. We are powered by SQLCoder – our state of the art open-source model that can search and visualise structured data (like SQL databases or Data Warehouses), and can be further fine-tuned and and deployed on-prem on your servers.

Team Size:5
Location:Mountain View
Group Partner:Brad Flora

Active Founders

Rishabh Srivastava

ML developer. Previously bootstrapped data APIs co to 100M+ monthly end users. I blog at rish.blog and am fairly active on Twitter (@rishdotblog)

Rishabh Srivastava
Rishabh Srivastava

Medha Basu

Co-Founder of Defog.ai. Former journalist/editor/content marketing person. Grew an enterprise content marketing firm from $0 to 7 figures in yearly revenue. Once interviewed a terrorist.

Medha Basu
Medha Basu

Company Launches

Hey there! We’re Medha and Rishabh from Defog.ai.


Defog embeds in your app and lets your users query data in seconds, using everyday language. It’s like ChatGPT for data - right within your app. Sign up to get early access. Or give it a try at defog.ai/ask.

😤 The problem

Insights features within apps take a long time to build and get right.

This is because SaaS founders and product managers spend most of their time on improving their core product. Insights features are rarely top-of-mind, even though they increase user stickiness.

Most teams tell themselves that they will eventually get to it. But they rarely do. The result is a hastily-built dashboard that provides limited insights to users.

🎉 Our solution

We make it super easy to let your users ask questions in natural language and instantly get data and insights on your app.

Defog embeds an intuitive Q&A-style interface for analytics in your app, and lets users query data in seconds.

Some examples of how companies can use Defog:

  • Fintech companies could provide deep insights based on users’ transaction history.
  • Video meeting tools could extract common complaints across customer calls and make them easy to discover.
  • Social media platforms could replace generic analytics dashboards with more specific metrics.

🪄 How it works

  1. Give us the metadata of your database schema (if using SQL), or just unstructured data corpuses (if using APIs or text corpuses)
  2. Add our iframe or Javascript plugin to your app
  3. (optional) set up visual styling for how charts should look on our app
  4. Let your users ask questions of the data, and get insights in seconds

We’re powered by large language models that can search and visualise both structured data (like SQL databases or data warehouses) and unstructured data (like text in call transcripts).

PS: We don’t compromise on privacy 🔐 Our model only needs access to your database schema to work, and not actual data about your customers.

⚡ Our asks