---
title: "Omar Essam El Sherbiny (@omaressam) | Hashnode"
description: "Head of Data at WalaPlus. Data platforms on Google Cloud, data governance, applied ML. Read the latest articles by Omar Essam El Sherbiny on Hashnode."
url: "https://hashnode.com/@omaressam"
---

@omaressam

# Omar Essam El Sherbiny

@omaressam · Cairo, Egypt · Joined October 2025

Head of Data at WalaPlus. Data platforms on Google Cloud, data governance, applied ML.

[Website](https://omaressam.web.app/)

### [An image classifier you can trust 80% of the time: confidence gating and Grad-CAM](https://omaressam.hashnode.dev/an-image-classifier-you-can-trust-80-of-the-time-confidence-gating-and-grad-cam?utm_source=hashnode&utm_medium=feed)

A 93.7% CNN on Fashion-MNIST becomes a 99.1% auto-tagger for 80% of images once uncertain predictions are routed to a person.

4 min read · [omaressam.hashnode.dev](https://omaressam.hashnode.dev/an-image-classifier-you-can-trust-80-of-the-time-confidence-gating-and-grad-cam?utm_source=hashnode&utm_medium=feed)

[Oct 30](https://hashnode.com/posts/an-image-classifier-you-can-trust-80-of-the-time-confidence-gating-and-grad-cam/6ae43152eed63f2096326905)

### [Five customer segments, and the one that holds 65% of the lost revenue](https://omaressam.hashnode.dev/five-customer-segments-and-the-one-that-holds-65-of-the-lost-revenue?utm_source=hashnode&utm_medium=feed)

K-means on six engineered features doubles the silhouette of raw service flags, and separates churn risk without ever seeing the churn label.

4 min read · [omaressam.hashnode.dev](https://omaressam.hashnode.dev/five-customer-segments-and-the-one-that-holds-65-of-the-lost-revenue?utm_source=hashnode&utm_medium=feed)

[Oct 30](https://hashnode.com/posts/five-customer-segments-and-the-one-that-holds-65-of-the-lost-revenue/6ae43124bd254b6b4836001e)

### [Churn prediction where the simple model wins: logistic regression against five rivals](https://omaressam.hashnode.dev/churn-prediction-where-the-simple-model-wins-logistic-regression-against-five-rivals?utm_source=hashnode&utm_medium=feed)

Six model families tuned on the same folds; the top three are statistically tied, so the readable one gets the job.

4 min read · [omaressam.hashnode.dev](https://omaressam.hashnode.dev/churn-prediction-where-the-simple-model-wins-logistic-regression-against-five-rivals?utm_source=hashnode&utm_medium=feed)

[Oct 30](https://hashnode.com/posts/churn-prediction-where-the-simple-model-wins-logistic-regression-against-five-rivals/6ae430f75545bb9a9b844cb4)

### [The year the Nile changed: change-point detection on 100 years of river flow](https://omaressam.hashnode.dev/the-year-the-nile-changed-change-point-detection-on-100-years-of-river-flow?utm_source=hashnode&utm_medium=feed)

Three methods agree on a break in the Nile's flow at Aswan around 1898-1900; knowing about it removes a forecast bias, not the noise.

4 min read · [omaressam.hashnode.dev](https://omaressam.hashnode.dev/the-year-the-nile-changed-change-point-detection-on-100-years-of-river-flow?utm_source=hashnode&utm_medium=feed)

[Oct 30](https://hashnode.com/posts/the-year-the-nile-changed-change-point-detection-on-100-years-of-river-flow/6ae430c9bf394b44cfa947ef)

### [Pharaohs of Egypt: a validated dataset from Wikidata, built like a data pipeline](https://omaressam.hashnode.dev/pharaohs-of-egypt-a-validated-dataset-from-wikidata-built-like-a-data-pipeline?utm_source=hashnode&utm_medium=feed)

SPARQL extraction, cleaning with date precision, pandera validation and a quality log: 513 pharaohs, 38 dynasties and 84 pyramids in three tables.

4 min read · [omaressam.hashnode.dev](https://omaressam.hashnode.dev/pharaohs-of-egypt-a-validated-dataset-from-wikidata-built-like-a-data-pipeline?utm_source=hashnode&utm_medium=feed)

[Oct 30](https://hashnode.com/posts/pharaohs-of-egypt-a-validated-dataset-from-wikidata-built-like-a-data-pipeline/6ae43095bd254b6b4835a691)
