Esther Kituyi | Engineer & Data Scientist

01.

About Me

Hello! My name is Esther, and I enjoy creating things that live on the internet and in the physical world. My interest in engineering started when I realized that power systems could be smarter, more efficient, and data-driven.

Fast-forward to today, I hold a BSc in Electrical & Electronics Engineering and I am a Fellow in Data Science. My main focus these days is building accessible, inclusive products and digital experiences at the intersection of Hardware (IoT) and Software (ML).

Here are a few highlights of my journey:

Smart Grids
Python & SQL
IoT Prototyping
Anomaly Detection

Education

BSc. Electrical & Electronics Eng.

Data Science Fellow

03.

Selected Projects

Tanzania Tourism Prediction

RANK #1

REGRESSION & ENSEMBLE

Achieved Rank #1 on the Zindi leaderboard! Predicted tourist expenditure to assist the Tanzania Tourism Board using XGBoost, intense feature engineering, and strategic outlier retention.

XGBoost Regression Feature Eng.

SuperLender: Default Risk

CLASSIFICATION & SCORING

Developed a credit scoring model to assess customer creditworthiness and calculate default risk. Overcame massive class imbalance. Core project for M-KOPA portfolio application.

Classification Class Imbalance Risk Modeling

COVID-19 Sentiment Analysis

CURRENT

NLP & MACHINE LEARNING

Analyzing vaccination sentiment by transitioning from hard classification to a continuous RMSE target (soft labels). Generating word clouds and interpreting complex confusion matrices.

NLP TF-IDF Tokenization

Smart Grid Anomaly Detection

DATA ENGINEERING & ML

Engineered a system using Python to identify technical faults and energy theft signatures in unstructured sensor data. Utilized Medallion Architecture for data pipelines.

Python Scikit-Learn ETL

Waste-to-Cash Validation

COMPUTER VISION

Designed to validate waste images for a "cash for trash" initiative. Developed using Convolutional Neural Networks (CNNs) and augmented through synthetic dataset generation.

CNN Computer Vision Synthetic Data

Air Quality Forecasting

TIME-SERIES & CLUSTERING

A Zindi Challenge focused on predicting future air pollution levels. Combined geospatial clustering techniques with time-series forecasting, rigorously evaluated on RMSE.

Time-Series Clustering Forecasting

Financial Inclusion in Africa

BINARY CLASSIFICATION

Predicting bank account ownership across Africa to improve financial accessibility. Relied heavily on deep demographic data analysis and robust binary classification modeling.

Demographics Classification EDA

IoT Smart Metering System

HARDWARE & IOT

GSM-enabled prototype for real-time utility monitoring. This serves as a foundational project from my Electrical Engineering degree at DeKUT, focusing on energy optimization.

Arduino/C++ GSM Integration Data Collection

View More on GitHub

04.