as.

Data scientist Seattle, WA

Alhassan Sahad.

I work with data from forests, fires, and the everyday.

My work spans forestland monitoring, wildfire research, and machine learning on large datasets. I came to data science through geological engineering, and now build models at Weyerhaeuser. Previously, I worked with wildfire data at NASA.

Email me GitHub (opens in a new tab)
Alhassan Sahad
Geology → data science

01 / Projects

Selected work.

Projects in computer vision
and customer analytics.

Computer vision 2023

Finding similar images with VGG16

Can a photograph find the right shoes when words fall short?

A photograph becomes a set of features. Those features help find shoes with similar shapes and colors in a 300-image catalog.

Method
VGG16 embeddings · cosine distance · nearest neighbors
What I found
The saved searches retrieve similar shapes and colors. Searching a larger catalog remains a next step.
Read the project
Search by image

Choose a photo to see its saved search results.

Closest matchesRed heel
Eight retrieved shoes, mostly red heels, with their cosine distances above each image.
Closest matchesRain boots
Eight retrieved boots, including patterned rain boots, with their cosine distances above each image.

Two examples from the original experiment. Lower distance means a closer match.

Image classification 2023

Recognizing fruit with a CNN

Can a model distinguish fruit categories from photographs?

Convolutional neural networks · image augmentation

Takeaway The baseline learning curves show a gap between training and validation performance: training accuracy alone is not enough.

Read the project
Baseline training and validation curves

Customer analytics 2023

Finding patterns in grocery baskets

What shopping patterns emerge when customers are grouped by purchase behavior?

Behavioral features · normalization · K-Means

Takeaway Clusters are a starting point. Interpreting their purchase patterns is what makes the groups useful.

Read the project
Comparing the number of clusters

Predictive modeling 2022

Predicting customer loyalty

Which purchase and engagement signals help distinguish loyal customers?

Feature engineering · classification · model evaluation

Takeaway Purchase frequency and time between purchases helped characterize loyalty in this experiment.

Read the project
Purchase historyBehavioral featuresLoyalty prediction

02 / Experience

The path here.

From geological engineering
to models of the world around us.

2024 — Present

Weyerhaeuser

Data Scientist Seattle, WA

I build machine learning models to support the monitoring of more than 10 million acres of forestland.

2023 — 2024

NASA Marshall Space Flight Center

Data Scientist Huntsville, AL

I analyzed terabyte-scale wildfire datasets and built TensorFlow classification models to study wildfire behavior.

2022 — 2023

Institute for Data Science and Informatics

Data Scientist Columbia, MO

I built machine learning models for environmental data analysis at the University of Missouri.

Education

M.S. Data Science and Analytics

University of Missouri–Columbia 2024

B.S. Geological Engineering

Kwame Nkrumah University of Science and Technology 2019

03 / Writing

Notes on the work.

Making sense of models
and the data that goes into them.