Product research and PRDs
Turn an early idea into a clearer problem, user need, assumptions, constraints, edge cases, trade-offs, success criteria, and acceptance criteria. Sometimes the right answer is still: we should not build this.
Product architecture in training
I work between a business question and a clean engineering handoff
Marketing, business and growth background. I learn Product Architecture by getting close enough to code, Git, tests, docs, maintenance and AI-assisted development to see the constraints before they turn into expensive confusion.
142
Public and private repositories
4,869
Commit contributions
8,947
Contributions this year
81
Day contribution streak
Snapshot of the GitHub profile, committed on 2026-09-19.
- Product research
- PRDs
- Technical docs
- Bounded PRs
- Test design
- Agent workflows
- Evidence gates
- Arabic conversion
- AI evaluation
- Build in public
What I own
Six places I can take real responsibility
The useful part of this profile sits between commercial context and engineering constraints. These are the pieces of work I can carry end to end.
Docs and technical writing
Keep READMEs, onboarding, decision notes, architecture context, and product documentation close to the thing that actually exists, not the version everyone remembers from three months ago.
Issue discovery and bounded PRs
Reproduce problems, gather evidence, narrow scope, write useful issues, and take small changes as far as a reviewable PR when they are inside my technical range. Engineering keeps the final judgment.
Test design and acceptance scenarios
Translate product intent into user stories, user journeys, acceptance scenarios, test suites, and test cases that cover expected behavior, edge cases, regressions, and failure paths.
AI-assisted workflows and agentic tooling
Look at repeated work, decide what AI can prepare safely, define where human review belongs, and turn recurring research, checks, or procedures into reusable skills, workflows, or small internal tools.
Evaluation, cost, and context discipline
Inspect repeated prompts, unnecessary context growth, oversized models doing simple work, weak verification, and duplicated generation. Then test alternatives and measure before claiming savings or quality gains.
Builds
Systems getting my attention right now
Selected, not exhaustive. Each one is a public repository you can read, clone and challenge.
Get Fable (opens in a new tab)
A coding-agent harness built around research, planning, TDD, verification, review, red-teaming, recovery, and persistent context.
- planning
- TDD
- verification
Agent Kernel (opens in a new tab)
Memory, governance, trust boundaries, and failure lessons for coding agents across multiple hosts.
- agent memory
- governance
- trust boundaries
Riqor (opens in a new tab)
Evidence gates and session continuity for AI coding agents. No evidence, no ship.
- evidence gates
- session continuity
Delegate Team (opens in a new tab)
A CLI for routing bounded coding work across local agents with explicit task ownership and a reviewable shipping chain.
- multi-agent
- task routing
- CLI
Dokion (opens in a new tab)
A repository-owned hardening playbook that turns security and quality work into explicit, verifiable steps with rollback paths.
- hardening
- evidence
- rollback
Unslop Preflight (opens in a new tab)
Frontend quality gates that challenge generic UI decisions before an agent starts writing the implementation.
- frontend QA
- design gates
- preflight
Rafiq Bot (opens in a new tab)
An open-source multimodal AI companion for Egyptian Arabic with persona synthesis, stateful behavior, tools, and multi-agent orchestration.
- Egyptian Arabic
- multimodal
- persona engine
PyMC Marketing MCP (opens in a new tab)
Decision-safe Bayesian marketing science for agents: MMM, diagnostics, CLV, experimentation, budget allocation, provenance, and decision gates.
- Bayesian MMM
- MCP
- decision gates
WhatsApp AI Supervisor (opens in a new tab)
A model-agnostic AI supervisor and deterministic permission engine for WhatsApp Cloud API.
- WhatsApp API
- permission engine
- model agnostic
Omni Skill (opens in a new tab)
A cross-host skill compiler for ChatGPT, Codex, Claude Code, and Antigravity, built to make one skill definition portable across runtimes.
- cross host
- skill compiler
- portable
Marketing Council Pack (opens in a new tab)
A cross-agent marketing strategy pack with specialist skills, challenge gates, and evidence-aware decision workflows.
- strategy
- agent skills
- evidence gates
Designly (opens in a new tab)
A commercial art-direction plugin with 21 focused skills, bounded specialist agents, AI image and video direction, and independent visual QA.
- art direction
- AI video
- visual QA
LinkedIn Animated Infographics (opens in a new tab)
A cross-host visual storytelling workflow for animated 1080x1350 infographics with motion craft, RTL support, and visual QA.
- visual storytelling
- motion
- RTL
PatchPilot (opens in a new tab)
Turns ChatGPT Web into a PR-only GitHub agent. It inspects a repo, implements one focused change, opens a reviewable PR, and stops there.
- PR only
- review ready
- Python
Conversion Copywriting Engine (opens in a new tab)
Research, drafting, and copy-audit pipeline tuned for MENA markets and Arabic conversion work.
- conversion copy
- MENA
- copy audit
Stop-Slop (opens in a new tab)
Linter and pattern analyzer that removes predictable AI writing tropes from product and software copy.
- linting
- copy quality
- anti-slop
Stack
Working with and learning
Some of this I use daily. Some I am learning deeper through current builds.
Agents, models and protocols
- ChatGPT
- Claude
- Claude Code
- Codex
- Gemini
- MCP
- Agent Skills
- Cursor
- GitHub Copilot
- OpenAI API
- Anthropic API
Frontend and product UI
- React
- Next.js
- Vite
- Tailwind CSS
- shadcn/ui
- Zustand
- Astro
Backend and API
- Bun
- Node.js
- Python
- FastAPI
- TypeScript
- JavaScript
- REST APIs
- Serverless
- Zod
- Pydantic
Data, auth and storage
- Supabase
- PostgreSQL
- Prisma
- SQL
- Row Level Security
- IndexedDB
- Dexie.js
Testing and quality
- Vitest
- Playwright
- React Testing Library
- Pytest
- axe
- ESLint
- Type checking
- TDD
Automation and delivery
- Git
- GitHub
- GitHub Actions
- n8n
- Docker
- Vercel
- CI/CD
- VS Code
Marketing and measurement
- Agentic Marketing
- PyMC
- Meta Ads
- Google Ads
- GA4
- Google Tag Manager
- Search Console
- AEO / GEO
Method
How I work
Four rules that decide what gets built and what gets thrown away.
Problem before build
The starting point is repeated friction, never a tool category.
Context before automation
The system knows the audience, the repo, the rules, and the evidence before it acts.
Proof before claims
A test, a report, a PR, or a measured result carries the conclusion.
Human judgment stays close
Agents prepare, inspect, and verify. People own the consequential calls.
Background
Where I came from
I came into software from the commercial side: performance marketing, conversion, campaign strategy and digital leadership across MENA.
After years of asking technical teams to build things, I started building and maintaining my own tools so I could understand the other side of those decisions.
Founder of PrePilot, built around Arabic conversion work. Digital Director with a background in performance marketing, Meta and Google campaign architecture, and direct-response work across MENA. Earlier work included large-scale event marketing at the Hajj Conference and Exhibition.
Today I use public software projects to learn the technical constraints behind product decisions while continuing to work from the commercial context I already know.
A productive day
A vague issue becomes a reproducible issue with evidence. An early feature becomes research plus a brief. Stale docs get fixed. A small technical change becomes a bounded PR. A day does not have to end with a lot of code. It should end with something another person can inspect and use.
- PR
- Issue
- Doc
- Research note
- Decision
- Test evidence
- Skill
- Checklist
Author of PrePilot
PrePilot gives ChatGPT and Claude 526 structured agency workflows across strategy, paid media, SEO, AEO, GEO, content, ad copy, UGC scripts, landing pages, proposals, decks and reporting. It handles Arabic, English and mixed-language briefs, and keeps review, editing and approval with the team.
Open PrePilot (opens in a new tab)Questions
Common questions
What does Mamdouh Aboammar actually do?
He works between business questions and engineering handoffs: product research and PRDs, technical documentation, issue discovery and bounded pull requests, test and acceptance design, AI-assisted workflows, and evaluation of AI cost and context discipline.
Is he available for hire?
Yes. He is open to freelance, full-time and part-time work, across Egypt, Saudi Arabia, the United Arab Emirates and the wider GCC. Email mamdouhfces1997@gmail.com or message him on WhatsApp.
Is he a software engineer?
No. He comes from marketing, business and growth, and is learning Product Architecture by building, documenting, testing and maintaining real software in public. Engineering keeps the final technical judgment on his work.
What is PrePilot?
PrePilot is a marketing workflow product for agencies, freelancers and marketing teams that already work inside ChatGPT or Claude. It provides 526 structured agency workflows and supports Arabic, English and mixed-language briefs. Mamdouh co-founded it and authors its product and workflow direction.
Where can the code be reviewed?
All of it sits on GitHub under github.com/imMamdouhaboammar, with an interactive catalogue of 142 repositories and a full markdown catalogue organised into ten categories.