LLM document pipelines
Scanned PDFs go through page rendering, vision OCR and extraction with citations, then to a reviewer. Python workers on a queue, typed contracts, evaluation against gold answers.
I’m Abanoub. For 7 years I’ve built production software. Today I build LLM pipelines in Python and the Node.js APIs, React frontends and cloud infrastructure that put them in front of real users.
{ "name": "Abanoub Essam", "role": "Full-Stack AI Engineer", "based_in": "Ismailia, Egypt", "experience_years": 7, "languages": ["Arabic", "English"], "focus": ["LLM pipelines", "APIs", "system design"] }
Shipping production code since March 2019.
Cairo time. Remote-first, overlaps with Europe and the Gulf.
Public repositories on GitHub.
I build the Python services that turn scanned documents and open questions into structured, checked answers.
Vision-model OCR on scanned PDFs, then classification and field extraction into typed schemas.
Every answer links to the page it came from, so a reviewer can check it in one click.
Output is compared with hand-labelled gold answers before each release.
Async workers on Pub/Sub, idempotent jobs, structured logs and Docker deploys on Google Cloud. A person approves before anything is sent.
Grouped by layer. Every item has run in production.
Described by the problem and the stack. Client details stay private.
Scanned PDFs go through page rendering, vision OCR and extraction with citations, then to a reviewer. Python workers on a queue, typed contracts, evaluation against gold answers.
Chat APIs between a product and the model, with history, auth and rate limits.
Supplier search, booking, approvals and payments, with idempotent retries.
Role-based admin panels, customer portals, and the APIs behind Flutter apps, including moderation tools for user content.
Python LLM pipelines on Vertex AI, NestJS services and Terraform-managed cloud infrastructure.
NestJS APIs, React portals, and a FastAPI chat service built with a machine-learning team.
React and Next.js frontends on my own Node.js APIs. Mentored interns.
First production Node.js services. B.Sc. Computer Science, 2017.
Fixed-scope projects or a monthly retainer, in Arabic or English.
Document extraction, OCR, classification or chat added to your product, with evaluation so you know it works.
For: PDFs, support, internal toolsREST and GraphQL APIs, queues and caching that hold up under load.
For: new products, slow APIsDatabase to Next.js and Flutter: one engineer who owns the whole feature.
For: MVPs, dashboardsArchitecture reviews, code review habits, CI/CD and hiring help.
For: teams of 2–10 engineersSend a short note about the project, timeline and budget. Arabic or English.