Skip to content
moaz
MOAZ ESSAM · SOFTWARE ENGINEER · CAIRO

I find what's slow or broken, then build the fix.

I work in network operations at a telco and build software on the side: a coaching platform where every coach only sees their own clients, a browser tool my colleagues use to read router captures, and automations that took hours of manual work off my team each week. I use coding agents heavily, and I review everything they write.

TypeScript · React · Postgres · Supabase · Python
Claude Code · Codex · Cursor · OpenCode · Antigravity
raptor · npm run test:rls · real output, replayed
$ npm run test:rls
RUN v4.1.9
✓ tenant isolation — cross-tenant reads return 0 rows > coach A cannot read coach B…
✓ tenant isolation — cross-tenant reads return 0 rows > coach A cannot read coach B…
✓ tenant isolation — cross-tenant reads return 0 rows > client A1 cannot read anoth…
✓ tenant isolation — cross-tenant reads return 0 rows > anon sees nothing (deny-by-…
✓ positive controls — owners and coaches see what they should > coach A sees their …
✓ support_tickets (0110) tenancy > owner + admin read the ticket; other client, the…
✓ push queue (0104) — durable enqueue + batch claim + reaper (§2/§9) > a notificati…
✓ admin_actions (0108) — append-only audit ledger, admin-read only (§7.4) > the see…
✓ positive controls — owners and coaches see what they should > client A1 sees thei…
✓ positive controls — owners and coaches see what they should > client A1 can read …
… 331 more
Test Files 1 passed (1)
Tests 341 passed (341)
Duration 913ms
~
plays once when it scrolls into view · static under reduced motion
288 rls policies · migrations
58 tables, rls on every one · harness
341 cross-tenant assertions · ci
0 leaks · last run
SELECTED WORK · 04
01
Raptor
Three roles, one database, zero cross-tenant reads. Deny-by-default row-level security, shipped in the same migration as every table.
Expo Router · Supabase · Postgres RLS · LiveKit · 27 edge functions · LLM features behind review
300 commits / harness open source
02
crosstalk
Agents from rival vendors (Claude Code, Codex, Cursor) build one codebase as a team. A review finding is a claim to verify, not a command.
TypeScript · MCP · node-pty · CLI
318 commits / public
03
WireTrace
A single HTML file that decodes PCAPs and router logs in the browser, so support staff without Wireshark can triage. Nothing leaves the machine.
Vanilla JS · hand-written PCAP parser · DHCP, DNS, ARP, PPPoE · spoof detection
1 file, 0 deps / public source
04
Vault
One agent versus a supervised team, same game, same spec. What the team shipped and what it broke.
Three.js · Vite · crosstalk benchmark
364 vs 133 commits
HOW I WORK

The agent types. I decide.

Notice, prove, ask why, build, measure. Then the agent types and I decide. Spec first, review everything, tests decide. Claude Code for architecture and review, Codex for routine implementation, Cursor for in-file edits. Read the one time an agent got it badly wrong and how the harness caught it.

Read the process →
# CLAUDE.md · raptor (excerpt) ALWAYS ship RLS in the same migration as the table ALWAYS run test:rls before opening a PR ASK before touching auth, payments, or policies NEVER use the service role key client-side NEVER store money as floats

Open to roles where the job is solving problems.

Full-stack, backend, automation, tooling, data. Cairo · remote-friendly · English and Arabic.