Information Engineering @ University of Padova.
I indulge my self in systems design, monitoring and rewiring. I enjoy writing C Rust and Bash. My current interest revolves around systems softwarelogs, streaming and concurrency.
Daily driver is Arch + Hyprland + Neovim. I like to break and re-break my system. FOSS apostle.
Currently building an adaptive log sampler. It learns the normal behavior of a log stream using Markov chains over log templates, then uses KL divergence to decide what to keep when ingestion spikes. The goal is to build it small enough to run on a Raspberry Pi or other edge computing options.
The repos below are the pieces it's built from. Each one shipped on its own first.
trafilo — streaming framework in C. Per-bucket sharding, bounded queues, sliding windows, dispatcher pattern. The substrate the rest of the pipeline runs on.
dRain — online log template miner in Rust. Fixed-depth parse tree, one-pass matching. Collapses raw log lines into stable template IDs.
markov-rs — transition model + KL-divergence baseline learner in Rust. Takes template IDs, learns the normal transition distribution, flags drift.
Flow State — A conn.log wrapper and hasher for markov-rs.
HiveParser — multithreaded log parser in C, built on trafilo. Helgrind-clean.
- Site & writing — basz_
- Email — basliel.gugsa25@gmail.com
