Data Scientist

Education

• M.S., Business Analytics

  Emory University (May 2024)

• M.S., Information Technology

  Carnegie Mellon University (May 2023)

• B.S., Software Engineering

  Adventist University of Central Africa (November 2021)

Work Experience

Machine Learning Engineer Intern @ The Rwanda Space Agency (Summer 2022)

• Conducted exploratory data analysis on geospatial images to monitor and quantify deforestation and forest degradation in remote areas of Rwanda.
• Optimized sequence deep learning models (RNN, LSTM, GRU) for image segmentation to achieve higher accuracy and precision on geospatial images.
• Collaborated with team members to develop comprehensive reports and engaging presentations to effectively communicate project findings and insights to stakeholders and team members.

Cofounder & Data Lead @ Extra Technologies Ltd (June 2018 - July 2021)

• Developed and led the implementation of a cooperative management platform that increased efficiency in daily operations for over 16 cooperatives with 5000 members, resulting in a 15% reduction in overhead costs.
• Created analytics dashboards using PowerBi to track and analyze member performance for Tea Cooperatives.

Projects

Frame-Level Speech Recognition using Multilayer Perceptron

code

Overview

Developed an advanced frame-level speech recognition system using a Multilayer Perceptron (MLP). This project focused on transforming raw Mel Frequency Cepstral Coefficients (MFCCs) into precise phonetic transcriptions at the frame level.

Key Responsibilities

• Engineered an MLP model to learn feature representations and establish a non-linear classification boundary for speech recognition.
• Employed cross-entropy loss to minimize the discrepancy between the predicted outputs and the target phoneme class labels.
• Utilized gradient descent algorithms for neural network parameter optimization, aimed at reducing the overall cost function and enhancing model precision.

Achievements

• Successfully developed a neural network capable of distinguishing between diverse phonetic sounds in speech.
• Achieved a high model accuracy with a score of 0.87693, demonstrating advanced proficiency in speech processing and machine learning techniques.

Face Classification & Verification using Convolutional Neural Networks

code

Overview

• Implemented a ResNet architectures for face verification, focusing on constructing effective convolutional neural networks and generating discriminative, generalizable feature representations.
• Implemented a face classifier to extract and analyze facial features (like skin tone, hair color, nose size) from images, converting them into fixed-length feature vectors or face embeddings.
• Designed a verification system to compute the similarity between feature vectors of two images, determining if they represent the same person.

Key Achievements

• Successfully trained a model that can distinguish facial features with high accuracy, thereby enhancing the performance of face verification tasks.
• Developed and optimized a model capable of effectively comparing feature vectors to generate a similarity score, crucial for accurate identity verification.