Nader Mohamed

Data Scientist & AI Engineer

I train models and put them where the user already is.

One of them reads 171 hieroglyph signs at 98.2%a, inside a phone browser, with no photo leaving the device.

Sources

  • the project's own README, checked 2026-08-28

Work

Ten projects, in three groups. Each group makes a claim you can check against every project underneath it.

Models that leave the notebook

Trained, then put somewhere a person can open.

  • Wadjet

    Point a phone at a hieroglyphic inscription and it reads the signs back to you. The models run inside the browser, so the photo never leaves the phone.

    98.2% reported hieroglyph classification accuracy. Source: the project's own README. Checked 2026-08-28.

  • Incident Forecasting Pipeline

    Forecasting incidents across 129 locations. The best thing in it is a feature I deleted, because it would have made the model look perfect.

    0.8684 R² on the current dataset. Source: the public repo and the internal deliverable log, which agree. Checked 2026-08-28.

  • Cerebral Stroke Prediction

    The gradient descent is a loop I wrote, not an import.

Systems that stop

Every one of these knows a condition under which it refuses to continue.

  • Grounded RAG PDF Q&A

    Ask it something the documents do not answer and it returns one exact sentence. Making it shut up was the engineering.

    17/17 golden eval, twelve grounded questions and five adversarial ones. Source: the project's own README. Checked 2026-08-28.

  • workflow-doc-agent

    An engagement wanted an agent that documents production workflows. I built it before the call, took the call with it running, and got the work.

    9 tests, none of which touch the network. Source: the project's own README. Checked 2026-08-28.

  • Multi-agent reporting pipeline (client work, anonymized)

    Eleven stages, ten quality specialists that can fail a run, and a queue that drains overnight while nobody is watching.

    11 pipeline stages. Source: the delivered build, read directly. Checked 2026-08-28.

Second opinions

Each one measures the same thing twice, because one measurement is not evidence.

  • Paired time-series synchrony analysis

    7,991 paired measurements, run through two unrelated methods, then checked against each other. Delivered as four notebooks a researcher could open and run alone.

    7,991 paired computations, sixty-one channels across one hundred and thirty-one segments. Source: the delivered notebooks and their saved outputs. Checked 2026-08-28.

  • Sentiment Alignment, Amazon reviews against BERT

    A three-star review often reads positive. A five-star review often reads flat. This measures the gap instead of the rating.

  • Student Success Analytics

    Low attendance plus low forum activity predicted risk better than GPA. The students that rule misses have good grades and no engagement.

  • T2D Adipose Tissue Research

    Which genes changed is one question. Which of those changes matter is a question about the network.

About

I live in Desouq, a town in Kafr El Sheikh, up in the Nile Delta. Cairo time, GMT+2.

I am reading for a BSc in Artificial Intelligence at Kafr El Sheikh University. I started in September 2023 and expect to finish in July 2027.

Most of my work is the unglamorous half of machine learning. Cleaning data. Engineering features. Checking whether a number means what it looks like it means. Then getting the result somewhere a person can open it.

Wadjet is the clearest example. I trained three image classifiers, exported them to ONNX, and ran them client-side through ONNX Runtime Web. The photo never leaves the phone, and inference costs the server nothing, so it survives on a free tier. A model trained in a Delta town ends up running inside a browser somewhere else. That is the part of this I find worth doing.

I went through the DEPI Data Science track from 2024 to 2025, led the cohort team, and finished with Best Project and Best Member. The following summer I did two internships, one at the National Telecommunication Institute and one at Egypt's Ministry of Communications and Information Technology.

Since July 2025 I have been a Data Science intern at Zetta Global, forecasting operational incidents across 129a locations. Real data behaves differently from a downloaded dataset. It grows, the distribution shifts underneath the model, and the honest number goes down rather than up.

Since October 2025 I have been Head of Programming at IEEE Kafr El Sheikh, where I plan the curriculum and teach the Python sessions. Before that I ran the C++ track. I have competed in IEEEXtreme, the 24-hour global programming competition.

I freelance as well. Three completed contracts, two of them rated 5.0b.

What I am aiming at is production systems rather than research. I graduate in July 2027.

Sources

  • the public incident-prediction repo, checked 2026-08-28
  • Upwork profile, read directly, checked 2026-08-28

Now

Building
Extending the Zetta forecasting pipeline into driver risk scoring and segmentation, reading from and writing to a PostgreSQL warehouse. Also building this site.
Learning
Nothing new to report this month.
Available for
Data science and AI engineering roles, and freelance work.

Updated 2026-08-28.

Contact

Write to me at the address below. There is no form, because a form can break quietly and an address cannot.