paulopacitti

research

Mon, 01 Jan 0001 00:00:00 +0000

Research

My main area of research is cryptography, but I also study AI and computer science in general, so I might write new things about those. For now, these are my articles as an author and minor contributions with friends:

How does reducing the dimension of feature vectors impact Biometric Systems that use Homomorphic Encryption? (2025, minor contribution)
Ascon on 64-bit RISC-V: Software implementation on the Allwinner D1 processor (2024)
Building Trust with End to End Encryption: An Introduction to the Signal Protocol (2023)

reading

Mon, 01 Jan 0001 00:00:00 +0000

Reading

Here I store precious pieces of the internet (and academic papers) that made me grow in some way and helped thru my life. I really recommend reading of them:

AI/ML

“Training Compute-Optimal Large Language Models” (2022, Google) : the famous “Scaling Laws” paper, where researchers state that model size ($N$) and training data tokens ($D$) should scale equally under a fixed compute budget ($C$), resulting in a “LLM training formula” where you can even predict the training loss given these parameters.
“The Llama 3 Herd of Models” (2024, Meta): tech report on how Meta trained the Llama 3 models. It’s detailed and easy to read. It helps me a lot understanding the whole LLM training pipeline and process.

Systems Programing

“Git at any scale” (2026, Cursor): how Cursor designed their own git hosting platform to have better availability than GitHub (LOL).

Mon, 01 Jan 0001 00:00:00 +0000

this is the third iteration of my website. One I made with React, other React + MDX, but editing was not that fast enough to me. So I rewrote again lol here’s how this website is made:

hugo
a modified version of hugo-bearblog