Peter Avenant (@PeterAvenant) profile picture

Peter Avenant

@PeterAvenant

Head of Data & AI Engineering at https://nitter.cf/t.co/jnFGXAsKGe

deltavault.ai
Joined September 2013
  • Tweets2K
  • Following152
  • Followers330
620Media
Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)Photo by Peter Avenant (@PeterAvenant)
Mapping every column to a business attribute before a table can ship rebuilds a canonical model's cost, one column at a time. AIMS makes mapping mandatory only for critical columns and tracks the rest as coverage that warns, not blocks. deltavault.ai/blog/every-met…
Illustration for the article Every method stops at the concept

ALT Illustration for the article Every method stops at the concept

11
A Data Vault hub or link no longer has to be a table. It can be a statement an agent reads. Which stays a table for you?
14
Databricks, Snowflake and Microsoft Fabric each build their agent a semantic layer, so you write the same business meaning on every platform. AIMS writes it once, on no platform, and every agent reads the same data contract. deltavault.ai/blog/what-the-…
Illustration for the article What the agent reads

ALT Illustration for the article What the agent reads

1
27
Bronze shipped a year ago and gold still isn't trusted. That gap isn't a missing platform feature. It's six pieces of integration work nobody scheduled, and the first is one meeting to agree what a customer is. deltavault.ai/blog/six-steps…
Illustration for the article Six steps to your integration layer: from bronze to trusted gold

ALT Illustration for the article Six steps to your integration layer: from bronze to trusted gold

5
Data Vault wrote business meaning into tables because a table was all a machine could read. Your model's new reader reads sentences. Audit history and parallel loading now live in persistent staging, and the meaning lives in your catalog. deltavault.ai/blog/modeling-…
Illustration for the article Modeling for Machines That Can Read

ALT Illustration for the article Modeling for Machines That Can Read

7
Agents rarely miss the join. They miss what the column means. Would an agent understand your column names today?
7
An AI-drafted description reads like one a steward wrote, so the next AI run treats it as fact. In DeltaVault's Map & Match, every suggestion arrives with a confidence score and a written reason, and a person decides where the bar sits. deltavault.ai/blog/context-f…
Illustration for the article Context for AI, context by AI: metadata needs a confidence score

ALT Illustration for the article Context for AI, context by AI: metadata needs a confidence score

5
Your silver layer can keep its sources' shape and integrate last. What are you integrating that nobody asked for?
9
Never let a governance rule block someone from saving metadata to a branch. Block a half-governed draft and people stop writing drafts, and the work moves to a spreadsheet. Gate generation, approval, and promotion to main instead. deltavault.ai/blog/metadata-…
Illustration for the article A Model You Can Diff: Metadata Version Control for Data Teams

ALT Illustration for the article A Model You Can Diff: Metadata Version Control for Data Teams

2
Gold can read silver before anything is integrated. Keep each source table in its own shape, and add a key mapping and golden record beside it only when a named consumer needs one record. Which of your integrations had that consumer?
Illustration for the article Integration by exception

ALT Illustration for the article Integration by exception

11
Data Vault wrote business meaning into hubs and links because tables were all a machine could read. A language model now reads the meaning directly: 16 percent correct on a raw schema, 54 percent over an ontology of the same tables. deltavault.ai/blog/you-dont-…
Illustration for the article You don't need Data Vault anymore

ALT Illustration for the article You don't need Data Vault anymore

14
Data Vault builds a hierarchical link table for a customer's parent account. AIMS adds one parent key column and writes the rest down as a hierarchy relationship your AI agent reads. We translated every construct the same way. deltavault.ai/blog/what-a-hu…
Illustration for the article What a hub becomes when an agent reads the model

ALT Illustration for the article What a hub becomes when an agent reads the model

14
Nothing about the data moved, yet a language model went from 16% to 54% accuracy on business questions once an ontology described the tables. Your meaning no longer needs a table per concept. It needs a sentence in your catalog. deltavault.ai/blog/the-econo…
Illustration for the article The economics of meaning flipped

ALT Illustration for the article The economics of meaning flipped

8
Portability has to live somewhere. Put it in the generated SQL and you get code that runs everywhere and uses nothing well. Put it in the model and Databricks gets a real Lakeflow bundle. Change the target, not the meaning. deltavault.com/blog/portabil…
Illustration for the article The model is portable. The code is native.

ALT Illustration for the article The model is portable. The code is native.

1
25
Silver layer modeling methods tell you which business thing a table is about. Only the canonical model says what each column means, and it rewrites every source to do it. Your AI assistant needs that meaning to stop guessing. deltavault.ai/blog/every-met…
Illustration for the article Every method stops at the concept

ALT Illustration for the article Every method stops at the concept

7
If nobody wrote down what a code means, your AI agent will guess it for you. Who on your team approves that meaning?
10
Written as an ontology over the same tables, business meaning lifted large language model accuracy from 16% to 54%. Meaning used to cost a table per concept. Now it costs a sentence mapped to a column. Where does your team keep it today?
Illustration for the article The economics of meaning flipped

ALT Illustration for the article The economics of meaning flipped

7
We recorded our own AI losing to a data engineer. Not on a hard problem: one view over two staging tables. It inner-joined across staging layers and flattened the history the view exists to keep. Catching that in a second is data engineering, not prompt engineering.
Illustration for the article This is why we need data engineers

ALT Illustration for the article This is why we need data engineers

12
Your data model lives in a load notebook, so every migration sends someone back to reinterpret it. Keep the model in the catalog as governed metadata and render native Databricks code from it. A new target changes the template, not the meaning. deltavault.com/blog/keep-the…
Illustration for the article Keep the model stable, change the template

ALT Illustration for the article Keep the model stable, change the template

19
Data Vault wrote business meaning into table structure because a table was all a machine could read. An AI agent reads sentences now, so that meaning belongs in your catalog. How many of your tables exist for something other than meaning?
Illustration for the article Modeling for Machines That Can Read

ALT Illustration for the article Modeling for Machines That Can Read

6
Load more