~/home
Youssef Safeh
I build software end to end and think like the attacker trying to break it.
- ◆ build
- applied AI · systems · backend
- ◇ break
- security · fraud · access control
Software engineer and CS student. For two-plus years I've built production software and then tried to break it: applied AI and LLM tools, backend and cloud, and the security work around them. The proof is specific: a pipeline that cut one client's fraud losses by 31.6%, and the backend for a live game with 1,000+ players.
● building: Perimeter · external attack-surface monitoring
youssef@safeh: ~
Youssef Safeh · software engineer & security-minded builder
this is an interactive shell. type a command or click one.
youssef@safeh:~$ help
type a command, or click it:
whoami short bio
social links
clear clear the screen
ls ~/selected-work
all projects →- PerimetersecurityExternal attack-surface monitoring that treats consent as a security control: no verified ownership, no active scan.read the breakdown →
- Server-Authoritative RBACsystemsA permission system for a live game that treats the client as hostile: it requests, the server decides, and every decision is logged.
- Fraud DetectionsecurityA scoring pipeline for a client's e-commerce checkout that cut fraud losses by 31.6%, built to route decisions to a human, not to auto-punish customers.
- Applied AI / LLM SystemsaiAgentic and retrieval-augmented tools built on both local and API models, including a RAG stack that runs fully on-device.
youssef@safeh:~$ ./contact.sh --hiring
Open to software, security, and applied-AI roles.
If you want an engineer who ships the thing and then tries to break it, let's talk. The fastest reply is email.