The Auromake Blog
In-depth technical guides and architectural patterns for the modern engineer.
Latest Insights
Quantitative Evaluation Strategies for RAG Systems
Addressing the non-deterministic nature of LLMs requires a shift from binary unit tests to probabilistic evaluation pipelines using Ground Truth datasets and Model-Based Grading.
Building a Production RAG Pipeline
Building a chatbot is easy. Keeping it accurate is a data engineering nightmare. Here is how we solved the 'Stale Context' problem.
Escaping 'ClickOps'
The AWS Console is a trap. Learn why treating infrastructure as software is the only way to build scalable, reliable data platforms.
November 2025
Managing Concurrency in ELT Pipelines
Naive ELT patterns often fail in production due to aggressive locking. We analyze the architectural trade-offs between `TRUNCATE` and Incremental Deletes in high-availability environments.
Why We Migrated from Airflow to Dagster
Airflow cares if a task ran. Dagster cares if the table exists. Why this distinction fixed our broken dashboards.
Data Contract Enforcement
Why data teams shouldn't be the last to know about breaking upstream schema changes. We push quality left using data contracts.