Noesis
A private research assistant that answers from your own English and Arabic documents, checks every answer against its sources, and turns what it reads into training data.
- Python
- FastAPI
- SQLite
- RAG
- Gemini
Systems I designed and built, from the problem to the result.
10 projects
A private research assistant that answers from your own English and Arabic documents, checks every answer against its sources, and turns what it reads into training data.
A local-first studio where an AI agent plans, writes and verifies conversational training data grounded in real sources, then hands you a clean dataset in the format your training stack reads.
An adaptive web-scraping engine that tests every site before touching it, picks the lightest method that sees the real content, and hands back clean, de-duplicated text.
Know, with evidence, whether a website really works: layered DNS, TCP, TLS, HTTP and real-browser checks for one link or thousands, with a verdict, a confidence score and the reason for every one.
A website quality console that crawls a site in a real Chromium browser and turns SEO, accessibility, performance, frontend, network, security and privacy problems into organized, evidence-backed findings with suggested fixes.
A privacy-first ATS resume checker that scores a CV with transparent rules, matches it requirement by requirement to a job post, and writes cover letters grounded only in what the CV proves.
An AI interview simulator that generates a timed technical exam for any role and seniority, then grades it out of 100 with deterministic scoring, rubric-based AI evaluation and a personal study plan, in English or Arabic.
An AI-powered storefront where a Gemini shopping assistant, grounded in the store's own catalog through local RAG, recommends only real products, in English or Arabic.
QR table ordering for an Amman hummus-and-falafel house: guests order from their phone by tapping or by chatting with an AI waiter that speaks Jordanian Arabic, while staff watch every table live on the restaurant's floor map.
A 229-dish rooftop restaurant run from the table: guests scan a QR, order by tap or by chatting with an English and Arabic AI waiter, and staff see every table move live on the floor map.