Altimate AI (@AltimateInc) header
Altimate AI (@AltimateInc) profile picture

DataPilot - AI Teammate for Data Teams. The team behind #dbtpoweruser vscode extension

Sunnyvale, California
Joined February 2024
Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)Photo by Altimate AI (@AltimateInc)
Your agent's dbt change passed in dev. Of course it did. The dev schema holds a sample, your own role owns every object, and no dashboard reads from it. Production changes all three the day the change merges.
1
5
Second: a Snowflake role inherits every privilege of the roles granted to it. An agent on a developer's role can end up writing to production. Give agents their own read-only role, and their own warehouse with a 600-second timeout.
1
None of that needs a new tool. Does your agent run on its own role, or on yours?
Enterprise dbt lineage is too big to read by eye. Click "chat about this model" and ask for its upstream dependencies. You get a dependency diagram and the lineage detail. marketplace.visualstudio.com…
12
Before you save a Snowflake cost ownership rule, click refresh. You see the cost it would attribute to the team, and any conflict with another team, with the overlap in dollars. app.snowflake.com/marketplac…
4
Profile every CTE in a dbt model in one click. Each CTE runs on its own and reports its runtime and row count. One took 4.4 seconds for 99 rows. Now you know where to look. marketplace.visualstudio.com…
6
Your Databricks invoice gives you one number and names nothing you can act on. In one worked example, the month came to about $49K, and one serverless SQL SKU was about $18K of it.
1
3
Average CPU won't tell you a cluster is right-sized, because a hot cluster can be retrying failed jobs. How does your team find the resource behind a Databricks spike?
1
2
Streamlit apps run their queries on Snowflake warehouses, so their cost hides in the warehouse bill. Tie it back to each app: runs, cost, execution time and warehouses, sorted by cost. app.snowflake.com/marketplac…
7
Reviewing a dbt model's SQL usually starts with writing a long prompt. With Profile Query you run the model and click one button. Altimate Code writes the prompt and runs the review. marketplace.visualstudio.com…
6
Set an alert on Snowflake Cortex credits in about a minute. Pick the AI service category, choose Cortex Functions credits, and alert when credits exceed five over the previous day. app.snowflake.com/marketplac…
12
We check one thing before the markers, which is the confidence indicator. If it reads Low Confidence at 50%, narrow the question before you act on the answer.
1
An AI tool gives you a Snowflake cost number. Finance asks where it came from. What do you show them? In Altimate Studio, every number with a source gets a marker. Click it and you see the source table and the SQL behind it.
1
7
Studio adds those markers in a separate step, after it writes the answer. So the wording never bends to fit a citation. A number with no marker is commentary, and it stays out of the finance review.
1
3
Compiled dbt SQL is still hard to read. Right click, Visualize SQL, and the query becomes a diagram: every CTE, group by and join, with the code behind each node. marketplace.visualstudio.com…
13