Writing

A working log, not a magazine.

I write here for three reasons:

1. Project logs. Mostly around vague — design choices, what worked, what didn't, what the numbers actually say. Posting in public keeps me honest.
2. Paper reviews. Short summaries of probabilistic ML, LLM memory, and fairness papers I'm reading. One paper, one post, one takeaway.
3. Technical notes. Production lessons from data platforms, agentic systems, and the parts of ML that nobody writes about (reconciliation, data quality, audit trails).

Posts default to English. Some may have an Italian counterpart; most won't.

First posts coming soon.