Python · SQL · data quality · reporting

Analysis that shows its work.

I’m Ekow Yalley, an operations-minded data analyst and University of Michigan graduate with a B.S. double major in Data Science (Econometrics) and Biology. I use Python, SQL, and statistical modeling to turn operational questions into clear analysis.

Selected case studies

Start with the decision, then inspect the evidence.

The featured work combines reproducible public projects with professional analysis built from private operating data.

01
Predictive modelingPython + AWS

AWS NYC 311 Service Resolution Modeling

Can public service requests be classified by whether they close within 72 hours—without leaking the answer into the model?

  • 78.71% accuracy
  • 79.88% balanced accuracy
  • 88.30% ROC AUC
  • 6/6 validation checks passed
ROC curve and confusion matrix for the NYC 311 service-resolution model
ROC curve and confusion matrix generated from 10,000 saved test-set predictions.
02
Reporting pipelinePython + pandas

Regional Sales Reporting Pipeline

How can two small source files become a clean, repeatable regional report with validation before and after the join?

  • $7,816 source-to-output reconciliation
  • 8 pipeline checks
  • 4 unit tests
03
Professional analysisPrivate data

Rail Operations Data Exploration

As Data Lead, structured car-, condition-, and segment-level comparisons across months of train-run data and turned recurring behavioral differences into visual material for technical team discussions.

Project library

What I explore in my free time

Games, numerical experiments, and small technical builds that let me test ideas outside formal project work.

Fun project · runnable

Me vs. You

A two-player Snake game built with Python Turtle, including independent controls, scoring, growth, and player-to-player collision logic.

View project →

Computational methods · reviewed

Numerical Computing Experiments

Power iteration, Numba acceleration, Monte Carlo estimation, and recorded HPC timing experiments.

View project →

How I work

Trust is part of the deliverable.

01

Define the question

Translate a broad request into a specific decision, target, or reporting output.

02

Audit the inputs

Check types, missing values, duplicates, join keys, ranges, and row-level exceptions.

03

Build reproducibly

Keep setup, transformations, dependencies, and run instructions understandable to another analyst.

04

Reconcile the result

Compare totals, test assumptions, quantify exclusions, and state what the output cannot prove.

Working toolkitPythonSQLpandasscikit-learnMySQLAWS S3AthenaJupyterGit

About Ekow

Operations-minded analyst, startup builder, and nonprofit founder

I’m most interested in the point where numbers become a practical decision: what to prioritize, where a process is breaking down, or how to make a recurring workflow clearer.

At Eco-Rubber Limited, I build data workflows spanning agricultural planning, factory monitoring, inventory, purchasing, and logistics.

I also founded Equiduct Education Program, a 501(c)(3) that connects college volunteers with K–12 students for free tutoring and mentorship.

Outside work, I’m a Manchester City supporter and NBA fan, and I follow both the Premier League and Champions League closely.