Latest blog posts by Andrzej Gabryel

How LLMs extract structure from unstructured data on Databricks: sampling, schema-enforced outputs, validation, quarantine tables and human review.

Custom agents, instruction files, skills and mandatory gates turn GitHub Copilot from an ad-hoc assistant into an engineering workflow you can audit.

Specialized AI agents keep implementation, testing, debugging and research in separate, reviewable lanes instead of one long coding assistant chat.

Learn to use AI as a junior data engineer to offload repetitive work, speed up pipelines, and focus your time on the decisions that actually matter.

AI now writes much of our SQL. What that changes about knowing data, as the hard part moves from writing queries to modeling, translation and doubt.

Manual data profiling slows the move from Bronze to Silver. Use AI-assisted rules and automated checks in Databricks to raise data quality faster.

Real-time data is a trade-off, not a goal. How to choose batch, streaming or micro-batch in CPG and retail, with POS, OSA and promo compliance examples.

Using GenAI for pipeline incident triage: an operating model for faster debugging, better runbooks and lower MTTR, without new security or on-call risks.

Most AI projects stall before production. Why data engineering teams should run a business-first Proof-of-Value instead of a Proof-of-Concept to show ROI.

An AI-powered operating model for data engineering turns noisy pipeline failures into fast, auditable recoveries, with earlier detection and safe fixes.