hello, world.|
I'm Aroa Xinping
Studying Data Science and specialising in FinOps. I like answering my own questions with data and building tools along the way.
/ projects
Real data and open source. Each one solves a specific problem with a real dataset.
cloud-cost-predictor
python · xgboost · plotly
123,357 real VMs, $5.9M/month wasted. Quantile XGBoost with asymmetric loss.
drive-failure-predictor
python · polars · xgboost
SMART data from 318,426 Backblaze drives. F1 0.77 on temporal validation.
leaked-key-impact
python · streamlit · gmail api
5 leaked AWS keys, 1,301 attempts in 33 days. Cost: $6,187/day.
room-occupancy-co2
python · jupyter · viterbi
Occupancy from a single CO2 sensor. Physical limit: yes or no, not how many.
vanguard-ab-test
python · sql · tableau
Vanguard A/B test: 70,609 clients, 755,405 web events.
gender-pay-gap
jupyter · pandas · ipums
IPUMS-CPS: 70.6%, 19.1% or 15.0%. Three figures, same data.
/ blog
Tools I discover, guides I wish I'd had, reviews of what I use and tech news.
/ about
Where I come from, how I got here, and where I'm heading.
Fine Arts → Dietetics → Data Science. Non-linear path, but every step contributed something.
Data Science (UOC)Data Analytics (Ironhack)Python · SQLFinOps
The direction is FinOps: how much of a cloud bill buys nothing, and which machines fail before anyone replaces them.
/ setup
What I've built, what I work on, and what I have fun with.
🖥
Custom Build PC
Main tower. Dual boot with CachyOS (Hyprland, terminal-first) and Windows 11.
CachyOS + Windows 11, Hyprland, Neovim
⌨
Homelab
Compact homelab running ZimaOS. Self-hosted services, backups and local network tinkering.
ZimaOS, Docker, Tailscale
💻
MacBook Air M1
For when I'm not at home. Light, silent and with battery for the whole day.
Apple M1, macOS, portable
🖨
Ultrawide
Where the PC and the Mac come together. One screen for code, dashboards and everything else.
Ultrawide, USB-C, KVM