// hello, stranger
Engineer @ Cloudflare

Alejandro
Sandoval.

sando · cdmx · aprendiendo en público

I work on AI security at Cloudflare. Right now I'm down the prompt-injection rabbit hole — how LLMs get tricked, and how to catch them failing quietly. Writing it all down as I learn, because I'm gonna forget otherwise.

LSPM ↗ prompt-injection 35mm cs50p learning in public
sando · 05.01.26
Alejandro Sandoval
cdmx · 19.43°N
education
B.S. Information Technology
UNAM · Mexico City
currently
Engineer @ Cloudflare
Making LLMs harder to fool
previously
Cisco · TrendAI
Network Support · Cybersec Intern
Cisco TrendAI
throughline
Networks → security → ML. Same curiosity, different layer of the stack.
what i'm doing this week // week 18 of '26
reading
since apr 23
The Pragmatic Programmer
Hunt & Thomas · ch. 4
learning
phase 0 · ml roadmap
CS50P · week 3
loops & exceptions
building
v0.2 · live
LSPM
prompt-injection · 89% recall
this month, off the clock // may '26
analog
127frames
shot in '26 · 35mm, slow life
album / month
may '26
In Rainbows
Radiohead
open in spotify →
code
47commits
this month · shipping LSPM