Buster

Hey, this is Blake & Dallin from Buster.

We're building AI Digital Workers for every part of your analytics & data stack. You can think of our platform as a 24/7 team of AI data engineers & analysts.

Its modern, simple, & open source. You can watch a demo video here.

what does Buster actually do?

If you’re a data professional or engineer… Buster gives you a team of 24/7 AI Workers that monitor your data stack, suggest model improvements, & handle ad-hoc requests from users.

If you’re an operator (or really anyone else)… Buster serves as your personal AI Data Analyst. This AI Data Analyst can answer ad-hoc data questions, do deep dive analysis on your behalf, & build beautiful dashboards or reports for you.

problem

Companies struggle to effectively leverage AI to get more value from their data. There are a few reasons for this.

1. Data is siloed. By default, company data siloed away in various places. To solve this, teams usually ETL their data into a warehouse, then perpetually build ad-hoc views or dbt models. In other words… they become a data stack Frankenstein with poor governance - making it impossible for LLMs to safely work with their data.
2. Data modeling is broken. The typical modeling workflow is: write a model → publish it → query it. It’s linear & makes it hard to create a feedback loop from the BI layer. This results in building hundreds of new models instead of improving existing models.
3. Current BI tools aren’t code-based (more to come on this).
4. Current BI tools just aren’t great products. This is a pretty consensus idea.
5. LLMs weren’t performant enough, but now they are. Prior to GPT-4o & 3.5 Sonnet, SOTA models just weren’t capable of accurately accomplishing data workflows (like text-to-sql). But now we are seeing consistent performance across dozens of workflows (see this post from Garry Tan about our evals).

solution

Here are some things that make the Buster platform unique & reliable.

1. A single, shared data model. We integrate and unify all of your company data into a single data model. This approach provides a strong foundation for AI workers to accurately utilize your data.
2. Guardrails & AI safety. Lets say a user requests some kind of analysis, but it introduces a new concept that isn’t clearly defined in the data model. Buster will identify this & flag the request. Then, an AI Worker will automatically create a branch & generate a suggested model update. You can review & merge it’s suggested update with one click.
3. Everything in Buster is code-based. Your model, dashboards, metrics, documentation, permissions, etc all live (as files) in your own Github Repo. This enables you to manage everything from your CLI & CI/CD pipeline. More importantly, it enables AI Workers to do things like:
   • Automatically ingest metadata from other tools & build your data model
   • Automatically generate & push improvements to your data model over time
   • Document key business logic or terminology
   • Fix impacted dependencies when there are breaking changes
   • Utilize git & can deploy via CI/CD
   • Integrate seamlessly with dbt & communicate with other tools you use
   • And much, much more
4. Personal AI data analyst. End users can just tell Buster the exact metric, dashboard, or analysis they want to create & Buster will just… do it. It’s like v0 for business intelligence. And, “It. just. works.”

why us

Ultimately, we believe that the future of AI analytics is about helping engineers & data teams build powerful, self-serve experiences for their users. We think that requires a new approach to the analytics stack. One that allows for deep integrations between products & allows data teams to truly own the entire experience.

ask

Get started with Buster here. If you like what we’re building please star us on GitHub!