Also shipped in production: Laravel · CodeIgniter · Angular · RESTful APIs · SQL optimization — Tooling: Git · Docker · Figma

I design and ship LLM-orchestration systems — pipelines where multiple specialized agents, not a single prompt, produce work that has to survive scrutiny. My background spans backend engineering (Laravel, CodeIgniter, Angular), an MBA in IT & Communication (TAG-GDU), and several years of freelance academic consulting — building CVs, motivation letters, and application strategy for scholarship applicants under real committee-level pressure.

That last part isn't incidental. It's where ScholarForge came from.

I build and ship from Khan Younis, Gaza, under active infrastructure constraints. That fact shapes how I work more than how I talk about it: async-first workflows, disciplined scope, and shipping in increments that survive interrupted connectivity — not a caveat, an operating constraint I design around.

The problem: thousands of scholarship applicants submit AI-generated motivation letters that all read the same — "Since childhood, I have been passionate about..." Committees notice. Single-prompt generators make this worse, not better.

The approach: a 10-agent orchestration pipeline that reverse-engineers a scholarship's actual evaluation rubric, selects which parts of an applicant's story serve that specific application, generates documents against the rubric instead of a template, and puts every claim through a hostile-reviewer simulation before it reaches the applicant.

Where it actually stands right now — no rounding up:

Full architecture, agent specs, and roadmap are documented in the repo itself — I keep it honest there on purpose, including what's not built yet.

Why I'm building it: I've done this work by hand. Over the past several months I supported 30+ students through competitive scholarship applications — manually orchestrating multiple LLMs in parallel, comparing outputs, catching hallucinations, iterating until a document could survive committee scrutiny. ScholarForge encodes that process into software instead of repeating it by hand every time.

_{EOF · Basil Al-Farra · Khan Younis, Gaza}