Hazem, here's my honest read.
You're not confused because you lack direction. You're confused because you have too many viable directions and not enough signal on which one compounds fastest from where you're standing. So let me cut through that.
Your actual edges right now: you ship real things, you understand AI tooling at a workflow level most engineers don't, you have infrastructure chops, and you're in a low cost-of-living country which means international remote pay would change your life. Those are the variables we're optimizing around.
What it looks like: You stop treating the job market as something you're "watching" and start treating it as a 90-day campaign. You target US/Canada/EU remote-first companies hiring senior frontend or fullstack engineers - specifically ones building AI-powered products, because that's where your real differentiation lives.
Next 30-90 days:
-
Rebuild your resume around outcomes, not tasks. "Built a Chrome extension to 1,000+ users" matters. "Built AI block generation plugin under pressure in an unfamiliar stack" matters. "React/Next.js developer at Brainstorm Force" does not differentiate you.
-
Build a tight portfolio site - not a blog, not a wall of projects. Three case studies max, each showing the problem, what you built, the result.
-
Target 5-10 companies per week on platforms like Turing, Toptal, Arc, and direct applications. P pplrioritize Series A-C startups building AI products - they value people who actually know how to work with LLMs in production, not just people who've read the docs.
-
Practice system design and behavioral interviews. Your weakness isn't coding - it's probably the performance aspect of interviewing, and that's fixable with reps.
Upside: This is the fastest path to a real financial jump. We're talking potentially 3-5x your current comp. That money buys you time, optionality, and breathing room to explore everything else.
Risk: It takes real time and emotional energy to job hunt while working full-time. Rejection rate for international remote roles from Egypt is high - not because of skill, but because of timezone bias and hiring pipeline friction. You might burn 2-3 months before landing something.
Who this is for: Someone who needs financial leverage before anything else. Someone whose best creative work will come after they're not worried about stability.
My honest read on you: This is you. The financial jump would change your entire calculus on everything else.
What it looks like: You stop being a React developer who uses AI and start positioning yourself as someone who deeply understands how to build software with AI - the workflows, the tooling, the architectural decisions. You create content around this; you build small public tools that demonstrate it; you become known for it.
Next 30-90 days:
-
Start writing publicly - Twitter/X, LinkedIn, or a dev blog. Not "here's how to prompt ChatGPT" content. Actual deep dives: how you structured your Claude Code workflow with
todo.mdfor cross-session memory, how you built the Copilot Telegram bot and what you learned about remote agentic workflows, how you approach scoped prompts vs. letting the model run wild. This is content almost nobody is making well right now. -
Ship one small open-source tool that embodies this expertise. Something like a CLI that scaffolds AI-assisted development workflows, or a template for agentic coding setups. It doesn't need to be big - it needs to be specific and useful.
-
Engage with the AI engineering community. Reply to people like Simon Willison, Swyx, Thorsten Ball. Not for clout; because these are your actual peers in thinking and you'll sharpen your ideas.
Upside: This builds a compounding asset - reputation. It makes Path 1 dramatically easier (companies come to you). It opens consulting, speaking, and eventually product opportunities. The "AI workflow" space is still early enough that someone with real depth can claim territory.
Risk: Content takes months to compound. You might write for 60 days and feel like nobody's listening. It also doesn't pay rent - this is a long game that needs to run alongside something that does.
Who this is for: Someone who thinks in frameworks, enjoys teaching, and has genuine depth - not just surface-level opinions.
My honest read on you: You already do this naturally - sharing insights in Slack, explaining AGENTS.md effectiveness to colleagues. You just haven't pointed that instinct outward yet.
What it looks like: You build something, but you're ruthlessly selective about what. You're right that most products get cloned instantly now. The moat is no longer "I can code this and you can't." The moat is now one of three things: distribution (you reach people others can't), data (your product gets better with use in ways that can't be replicated), or workflow integration (you're embedded in someone's daily process so deeply that switching costs are high).
Next 30-90 days:
-
Don't start building. Start observing. Spend 30 days in communities where people complain about tools - indie hackers, specific subreddits, niche Slack/Discord groups. You're looking for problems where the pain is specific to a workflow, not a generic "I wish I had an app for X."
-
Look specifically at MCP server integrations, agentic workflow tooling, or developer experience tools. These are spaces where your infrastructure knowledge and AI depth give you a real advantage, and where the audience (developers, technical teams) will pay for something that saves them time.
-
When you find something, validate before building. Talk to 10 people who have the problem. If 3+ say "I'd pay for that today," build an MVP in a weekend using everything you know about fast AI-assisted development. If not, keep looking.
Upside: If you hit on the right problem, a small product can generate $2-5K/month relatively fast, and that changes your relationship with employment entirely.
Risk: Most indie products fail. The search for the right problem can take months. And if you're doing this alongside a full-time job, you have maybe 10-15 hours a week - which means you need to be surgical about what you spend that time on.
Who this is for: Someone with patience, tolerance for ambiguity, and the discipline to not build something just because they can.
My honest read on you: You have the skills for this, but I think doing it right now - before securing better income - would split your focus in a way that hurts both the product and the job search. This is a Path 1-then-Path 3 sequence, not a standalone.
What it looks like: You shift from being a frontend engineer who uses AI to someone who builds AI systems. Fine-tuning models, building RAG pipelines, training LoRAs, deploying inference endpoints. You've already been poking at this - the LoRA experiments, the mechanistic interpretability interest. This path means going all in on that transition.
Next 30-90 days:
-
Work through a structured ML curriculum. Not a full degree - something like fast.ai's practical deep learning course, combined with hands-on projects.
-
Build 2-3 portfolio pieces: a fine-tuned model for a specific use case, a RAG system with real retrieval evaluation, an agentic pipeline with tool use.
-
Start applying for ML engineer or AI engineer roles - these are the roles with the highest salary ceilings and the most demand growth.
Upside: Highest long-term ceiling. AI/ML engineers command premium salaries, and the demand curve is still going up. This also future-proofs you more than frontend work does.
Risk: This is the longest path. Your GPU (GTX 1650, 4GB VRAM) limits what you can do locally - you'd need to lean on cloud compute or Colab. The transition from "frontend dev who experiments with AI" to "credible ML engineer" takes 6-12 months minimum, and you'd be competing against people with formal ML backgrounds. Also - and I'll be direct - the market for junior ML engineers is actually quite crowded. The premium is at the senior level, and getting there takes time.
Who this is for: Someone who's genuinely drawn to the math and the systems, not just the applications. Someone with runway to invest in a transition.
My honest read on you: Your interest is real, but it's more on the application side than the research side. I think trying to become a full ML engineer would be a longer, harder path than leveraging your existing strengths. Your edge is in building products and workflows with AI - not in building the AI itself.
Path 1 first, Path 2 in parallel, Path 3 after.
Here's why:
Money solves your biggest constraint right now - time. Landing an international remote role at 3-5x your current pay means you can work normal hours, live well in Cairo, and have actual bandwidth for side projects. Everything else gets easier after this.
While you're job hunting (Path 1), start writing publicly about your AI workflow expertise (Path 2). This is maybe 3-4 hours a week - one solid post or thread. It costs almost nothing, it makes your job applications stronger (hiring managers will find your content), and it starts compounding immediately.
Once you're in a better-paying role and have some financial cushion - say 3-6 months in - that's when you start the disciplined search for a product to build (Path 3). By then you'll have more market exposure, more ideas from working at a better company, and more runway to experiment.
Skip Path 4 for now. Not because it's bad, but because it's not where your highest leverage is. You can always go deeper on ML later - and honestly, the way AI tooling is evolving, application-layer expertise might end up being more valuable than model-layer expertise for most of the market.
The thing you should start tomorrow - literally tomorrow - is rewriting your resume with outcome-driven bullets and setting up profiles on 3 remote job platforms. Not because the other stuff doesn't matter, but because this is the domino that makes everything else fall.
You don't have a direction problem, Hazem. You have a sequencing problem. And the sequence is: get paid properly first, build reputation in parallel, then build products from a position of strength.