AI-Augmented Data Engineering, with Maksym Karashchuk

No Hallucinations: AI That Ships

AI-Augmented Data Engineering, with Maksym Karashchuk

Jun 2, 2026 · 29m 51s

Chapters

01 · Show intro and the five stages of AI in a data team

1m 24s

02 · Why organisations slow down on adoption

3m 5s

03 · Maksym's day now versus a year ago

3m 51s

04 · Real agent examples: recordings vault, auto-documentation, the Azure DevOps loop

5m 7s

05 · Human in the loop: the engineer verifies the pull request

7m 4s

06 · The first time a machine opened the pull request for you

7m 58s

07 · Adoption among peers and the blockers

8m 40s

08 · Information overload and European governance on model inference

10m 50s

09 · The guess machine: what you cannot trust an LLM with

12m 56s

10 · Moving from wave one to wave two

13m 40s

11 · Organisational readiness, BPMN, humans-with-agents versus agents-with-humans

14m 15s

12 · The data warehouse homework an agent needs

15m 56s

13 · Infrastructure, backend, and frontend as code

16m 32s

14 · Generating project rules, garbage in garbage out

17m 24s

15 · Where to store rules: Markdown and token economics

18m 41s

16 · RAG versus Markdown, and why the vector database hallucinates more

19m 47s

17 · Version control for the knowledge base

21m 36s

18 · Treat rules like an architecture decision record

22m 13s

19 · The role shift: from code repository to knowledge base and agents

23m 23s

20 · Eighteen months out

24m 22s

21 · Better models or better processes, and running Sonnet 4.6

25m 51s

22 · Advice to data engineers: touch the technology, mind security

27m 4s

23 · Will the data engineer job exist in ten years

28m 29s

24 · Outro

29m 27s

Description

A data architect who builds agents every week gives an honest read on what AI is doing to the data engineering job. The job is not disappearing but it is...

A data architect who builds agents every week gives an honest read on what AI is doing to the data engineering job. The job is not disappearing but it is shifting: from writing code by hand to curating the rules, skills, and knowledge base that agents run on. As the LLMs are becoming commodity, the bottleneck in 2026 transitions form being a top model quality to the correctly built processes.

show less

Comments

Information

Author	Astral Forest
Organization	Astral Forest
Website	-
Tags	#ai #analytics #architecture #astral forest #cdo #data #data leadership #datalake #datawarehouse #engineering #model #no hallucinations #podcast #reporting #technology