Why "AIworthiness"?

In aviation, nothing flies because it seems to work. An aircraft is airworthy when it conforms to its approved design and is in a condition for safe operation — and that claim is backed by evidence: requirements, analyses, tests and traceability from the top-level safety objective down to the smallest part.

AI is now entering that world: predictive maintenance, flight-data analytics, decision support in the cockpit and on the ground, and large language models that help engineers navigate thousands of pages of certification specifications. Those systems will need to earn the same kind of trust. That is what I mean by AIworthiness.

What I will write about

• LLMs on aviation documents — retrieval-augmented generation (RAG) and fine-tuning on regulations and certification specs such as EASA CS-E, and how to make answers traceable to the source paragraph.
• AI for flight safety — learning from operational and occurrence data to anticipate risk.
• Assurance for machine learning — what verification, explainability and data quality mean when the “design” is a trained model.
• Lessons from production — a decade of building ML and MLOps pipelines in industry, from engine calibration data to IoT analytics.

Following along

Every post lives here first, and I share most of them on LinkedIn as well — the discussion usually happens there, so look for the Discuss on LinkedIn button at the end of each post.