Upenyu Hlangabeza

Upenyu Hlangabeza

Software Development | Machine Learning | Data Analytics | Data Science

GitHub LinkedIn

About Me

As a Machine Learning Engineer, I specialize in developing production-ready AI solutions and scalable data pipelines using Python, TensorFlow, PyTorch, and AWS. My expertise spans Natural Language Processing, predictive modeling, and advanced analytics, with notable projects including Electric Vehicle Population Analysis and Human Skeleton Analysis, achieving 73% accuracy in posture detection.

I excel in transforming complex data challenges into practical solutions through Power BI dashboard design and SQL-based statistical analysis, while maintaining a strong focus on business objectives. Beyond my technical work with CI/CD pipelines and containerized workflows, I continuously expand my skills through certifications in data analysis, business intelligence, and risk management, ensuring I remain at the forefront of industry innovations.

For a closer look at my projects and technical contributions, visit my GitHub. Lastly, to view certifications and projects, please click the desired project, then navigate to the bottom and click either "View Certificate" or "Visit Project." It will immediately take you to the respective project or certificate.

Tech Stack

PHP :
SQL :
Java :
Python :
Laravel :
Django :
Excel :

My Projects

Object Recognition Project Summary on CIFAR-100 Dataset **Problem Overview:** The project tackles the challenge of processing large volumes of image data, emphasizing the need for automated object recognition due to the impracticality of manual methods. **Solution:** Using the CIFAR-100 dataset, the project employs a Convolutional Neural Network (CNN) to automate image classification across 100 categories, aiming to improve object recognition accuracy. **Methodology:** - **Data Handling:** Images were aligned with labels and processed through a CNN model. - **Model Design:** The CNN included convolutional, pooling, and dense layers, using ReLU and softmax for classification. - **Training & Evaluation:** The model was trained over 10 epochs, and its performance was evaluated with a confusion matrix and accuracy score. **Results:** The CNN achieved strong classification accuracy, effectively identifying objects in the dataset. **Conclusion:** The project demonstrates the effectiveness of CNNs for object recognition on the CIFAR-100 dataset, providing a robust solution for image classification.

Object Recognition

Object Recognition

Object Recognition Project Summary on CIFAR-100 Dataset

**Problem Overview:**
The project tackles the challenge of processing large volumes of image data, emphasizing the need for automated object recognition due to the impracticality of manual methods.

**Solution:**
Using the CIFAR-100 dataset, the project employs a Convolutional Neural Network (CNN) to automate image classification across 100 categories, aiming to improve object recognition accuracy.

**Methodology:**
- **Data Handling:** Images were aligned with labels and processed through a CNN model.
- **Model Design:** The CNN included convolutional, pooling, and dense layers, using ReLU and softmax for classification.
- **Training & Evaluation:** The model was trained over 10 epochs, and its performance was evaluated with a confusion matrix and accuracy score.

**Results:**
The CNN achieved strong classification accuracy, effectively identifying objects in the dataset.

**Conclusion:**
The project demonstrates the effectiveness of CNNs for object recognition on the CIFAR-100 dataset, providing a robust solution for image classification.
Project Summary: Evaluating LSTM Networks for Gesture and Pose Analysis **Overview** This project assesses Long Short-Term Memory (LSTM) networks for predicting human gestures and poses, comparing them with Convolutional Neural Networks (CNNs) and traditional Recurrent Neural Networks (RNNs). It uses the KIMORE dataset, which includes 3D skeleton data from 78 subjects performing rehabilitation exercises for low back pain. **Techniques and Methods Implemented** **Data Representation:** Sequential 3D coordinates of human skeletal joints are used for time-series analysis, making LSTM networks particularly suitable. - **Data Processing:** - Normalization: Standardized joint positions to ensure consistency. - Feature Extraction: Extracted joint angles, velocities, and distances for enhanced model input. - Data Augmentation: Applied rotation, scaling, and mirroring to diversify the dataset. **Model Comparison:** LSTM networks were evaluated against CNNs and traditional RNNs to determine the most effective approach for gesture and pose prediction. This study was aimed to identify the best model for automated assessment of rehabilitation exercises.

Human Skeleton 3D Graph Data Analysis

Human Skeleton 3D Graph Data Analysis

Project Summary: Evaluating LSTM Networks for Gesture and Pose Analysis

**Overview**
This project assesses Long Short-Term Memory (LSTM) networks for predicting human gestures and poses, comparing them with Convolutional Neural Networks (CNNs) and traditional Recurrent Neural Networks (RNNs). It uses the KIMORE dataset, which includes 3D skeleton data from 78 subjects performing rehabilitation exercises for low back pain.

**Techniques and Methods Implemented**

**Data Representation:** Sequential 3D coordinates of human skeletal joints are used for time-series analysis, making LSTM networks particularly suitable.

- **Data Processing:**
- Normalization: Standardized joint positions to ensure consistency.
- Feature Extraction: Extracted joint angles, velocities, and distances for enhanced model input.
- Data Augmentation: Applied rotation, scaling, and mirroring to diversify the dataset.

**Model Comparison:** LSTM networks were evaluated against CNNs and traditional RNNs to determine the most effective approach for gesture and pose prediction.

This study was aimed to identify the best model for automated assessment of rehabilitation exercises.
The problem involves designing hydrostatic thrust bearings with the primary goal of minimizing operational power loss when subjected to axial loads. The design parameters affecting power loss include flow rate (Q), recess radius (R0), bearing step radius (R), and fluid viscosity (μ). Hydrostatic bearings employ pressurized fluid to create a thin film separating moving and stationary surfaces, reducing friction and wear while offering high stiffness and load capacity.

Hydrostatic bearing

Hydrostatic bearing

The problem involves designing hydrostatic thrust bearings with the primary goal of minimizing operational power loss when subjected to axial loads. The design parameters affecting power loss include flow rate (Q), recess radius (R0), bearing step radius (R), and fluid viscosity (μ). Hydrostatic bearings employ pressurized fluid to create a thin film separating moving and stationary surfaces, reducing friction and wear while offering high stiffness and load capacity.
I developed a ray tracing engine to demonstrate the capabilities of advanced rendering techniques. The project aimed to create photorealistic images by simulating how light interacts with objects in a scene. I managed to design and develop multiple objects that could interact with the light, focusing on implementing efficient algorithms to trace rays from the camera through each pixel. This involved calculating intersections with surfaces and accurately rendering shadows, reflections, and refractions. This experience deepened my understanding of computer graphics and the principles of light propagation and object visualization.

Ray Tracing

Ray Tracing

I developed a ray tracing engine to demonstrate the capabilities of advanced rendering techniques. The project aimed to create photorealistic images by simulating how light interacts with objects in a scene. I managed to design and develop multiple objects that could interact with the light, focusing on implementing efficient algorithms to trace rays from the camera through each pixel. This involved calculating intersections with surfaces and accurately rendering shadows, reflections, and refractions. This experience deepened my understanding of computer graphics and the principles of light propagation and object visualization.
Building a chat server was intended to demonstrate the capabilities of parallel computing. This project allowed me to create a real-time communication platform where users could interact seamlessly. It served as the backbone for exchanging messages, sharing media, and managing user presence, ensuring smooth and efficient interactions. Through this experience, I enhanced my skills in networking, server management, and application design while showcasing how parallel computing can improve the performance and scalability of applications.

ChatServer App

ChatServer App

Building a chat server was intended to demonstrate the capabilities of parallel computing. This project allowed me to create a real-time communication platform where users could interact seamlessly. It served as the backbone for exchanging messages, sharing media, and managing user presence, ensuring smooth and efficient interactions. Through this experience, I enhanced my skills in networking, server management, and application design while showcasing how parallel computing can improve the performance and scalability of applications.
### Project Aim: Analyzing Mental Health in International Students Using PostgreSQL **Aim**: This SQL-based project investigates how studying abroad impacts international students' mental health, focusing on whether the length of stay influences depression, social connectedness, and acculturative stress. Using PostgreSQL, the project analyzes data from a Japanese university to replicate findings that social connectedness and acculturative stress are key predictors of depression. ### Key SQL Projects in PostgreSQL 1. **Depression and Length of Stay** - **Objective**: Query depression scores (PHQ-9) to assess how they vary with students' length of stay. 2. **Social Connectedness and Stay** - **Objective**: Use SQL to evaluate the relationship between social connectedness (SCS scores) and stay duration. 3. **Acculturative Stress and Stay** - **Objective**: Analyze acculturative stress (ASISS scores) using SQL to see if stress changes with time abroad. 4. **Comprehensive Mental Health Analysis** - **Objective**: Combine all factors (depression, connectedness, stress) in SQL to find key mental health predictors. This project heavily utilizes SQL in PostgreSQL to provide insights into mental health trends among international students.

Telematics-Mental-Health-SQL

Telematics-Mental-Health-SQL

### Project Aim: Analyzing Mental Health in International Students Using PostgreSQL

**Aim**:
This SQL-based project investigates how studying abroad impacts international students' mental health, focusing on whether the length of stay influences depression, social connectedness, and acculturative stress. Using PostgreSQL, the project analyzes data from a Japanese university to replicate findings that social connectedness and acculturative stress are key predictors of depression.

### Key SQL Projects in PostgreSQL

1. **Depression and Length of Stay**
- **Objective**: Query depression scores (PHQ-9) to assess how they vary with students' length of stay.

2. **Social Connectedness and Stay**
- **Objective**: Use SQL to evaluate the relationship between social connectedness (SCS scores) and stay duration.

3. **Acculturative Stress and Stay**
- **Objective**: Analyze acculturative stress (ASISS scores) using SQL to see if stress changes with time abroad.

4. **Comprehensive Mental Health Analysis**
- **Objective**: Combine all factors (depression, connectedness, stress) in SQL to find key mental health predictors.

This project heavily utilizes SQL in PostgreSQL to provide insights into mental health trends among international students.
In this project, I looked at the market share and yearly growth of several treadmill suppliers, including Apex Athletics, Hercules Gear, Spartan Sports, Steel Power, and Summit Strength. I used Excel to analyze data on sales and supplier performance to understand how these companies are doing in the treadmill market. The analysis shows: Market Share Distribution: A clear picture of how much of the market each supplier controls, showing their standing compared to others. Yearly Growth Rates: An overview of how each supplier has grown over the years, helping to identify trends in their performance. Comparative Insights: Key trends and patterns that highlight the strengths and weaknesses of each supplier, which can help businesses understand the market better. Overall, this project provides useful insights for anyone in the fitness equipment industry, allowing them to make smarter decisions in a competitive environment.

Excel MarketShare Insights: A Comparative Analysis

Excel MarketShare Insights: A Comparative Analysis

In this project, I looked at the market share and yearly growth of several treadmill suppliers, including Apex Athletics, Hercules Gear, Spartan Sports, Steel Power, and Summit Strength. I used Excel to analyze data on sales and supplier performance to understand how these companies are doing in the treadmill market.

The analysis shows:

Market Share Distribution: A clear picture of how much of the market each supplier controls, showing their standing compared to others.

Yearly Growth Rates: An overview of how each supplier has grown over the years, helping to identify trends in their performance.

Comparative Insights: Key trends and patterns that highlight the strengths and weaknesses of each supplier, which can help businesses understand the market better.

Overall, this project provides useful insights for anyone in the fitness equipment industry, allowing them to make smarter decisions in a competitive environment.

Certificates

DataCamp

DataCamp

Intermediate Python

Intermediate Python

The Intermediate Python certification equipped me with a comprehensive understanding of Python’s data science libraries, including NumPy, pandas, Matplotlib, and Seaborn, alongside mastering object-oriented programming and algorithm optimization. This certification enhanced my ability to develop scalable, efficient solutions that drive real-world impact through machine learning, data analysis, and automation.

In practice, I’ve utilized Python in various capacities—from building machine learning models using libraries such as SciKit-Learn and TensorFlow to automating data processing tasks and creating data visualizations for strategic decision-making. For instance, during my work on Human Skeleton Analysis with Machine Learning, I applied my Python skills to preprocess and analyze data, resulting in a model with a 73% accuracy rate in posture detection.

My work ethic is driven by the continuous integration of new knowledge and the pursuit of optimized, automated solutions. I approach challenges with a solution-oriented mindset, always looking for ways to leverage Python’s capabilities to streamline processes, enhance efficiency, and deliver meaningful results.
Datacamp

Datacamp

AI Fundamentals

AI Fundamentals

In this AI fundamentals course, I learned how to identify various use cases across different AI sub-domains and gained a clear understanding of Generative AI, including its key terminology.

I developed the skills to construct effective prompts for generative AI tools and explored the ethical considerations associated with AI and generative AI solutions. Overall, I now have a solid grasp of both the practical applications and the ethical dimensions of working with AI technologies.
Datacamp

Datacamp

Data Literacy

Data Literacy

**Summary of Data Literacy Course Learnings**

During the Data Literacy course, I acquired a comprehensive understanding of analytics fundamentals, data visualization, and data storytelling. I am proficient in describing the various analytic approaches utilized by data professionals, ranging from summary statistics to complex models and hypothesis testing. This has equipped me with the ability to interpret analysis results effectively and use common analytic terminology with confidence.

In the realm of data visualization, I learned to identify and utilize various visualization types appropriate for different analytical contexts. I am adept at interpreting visualizations to support informed decision-making and understand the potential for visualizations to be used misleadingly. This knowledge ensures that I can both create and critically evaluate visual data representations.

Additionally, the course enhanced my skills in data storytelling. I can now assess the background of different audiences and tailor my communication strategies accordingly. I am capable of identifying the essential components of a compelling data story and integrating them with visualizations to convey insights effectively. This skill set allows me to present data in a manner that is engaging and tailored to the needs of diverse audiences.
Datacamp

Datacamp

Data Analysis in Excel

Data Analysis in Excel

Through the Data Analysis in Excel course, I developed advanced skills in data manipulation and analysis, particularly using PivotTables, logical functions, What-If Analysis, and forecasting tools. I learned how to efficiently extract valuable insights from complex datasets using PivotTables, enabling me to uncover hidden trends and make data-driven decisions. The course also introduced me to keyboard shortcuts, significantly enhancing my workflow efficiency.

In the domain of logical functions, I gained the ability to identify ideal customer profiles and design targeted marketing strategies. This included exploring customer segmentation techniques to optimize customer success, drive revenue growth, and support business expansion. My understanding of these intermediate logical functions has equipped me to make data-informed decisions that align closely with business goals.

The What-If Analysis tool provided me with the expertise to conduct scenario analysis, allowing me to evaluate how different variables impact sales outcomes. By mastering sensitivity analysis, I can now make informed predictions and develop strategies to navigate dynamic business environments effectively.

Finally, the course taught me how to use forecasting techniques to anticipate future trends and visualize this data for stakeholders. I learned to leverage moving averages and trendlines to communicate my findings clearly, ensuring that I can present data-driven insights that inform strategic planning and decision-making processes.
Microsoft

Microsoft

Develop dynamic reports with Microsoft Power BI

Develop dynamic reports with Microsoft Power BI

Power BI Data Analyst Skills

Data Retrieval & Performance Optimization: Proficient in connecting to various data sources (Excel, relational databases, NoSQL) and optimizing data retrieval for efficient reporting.

Data Transformation & Cleaning: Experienced in using Power Query for data cleaning, transformation, and loading, ensuring high data quality for analysis.

Semantic Model Design: Skilled in designing and managing semantic models, including implementing star schemas and optimizing data granularity for better report performance.

Measures & Calculations: Capable of creating and managing both implicit and explicit measures, along with calculated columns and tables, to enhance data modeling.

Report Design & Visualization: Adept at selecting appropriate visuals, designing intuitive report layouts, and applying advanced filtering techniques for user-friendly, data-driven insights.

Workspace & Model Management: Competent in managing workspaces within the Power BI service, distributing reports, scheduling model refreshes, and resolving connectivity issues.
Datacamp

Datacamp

Intermediate SQL

Intermediate SQL

Through the Intermediate SQL certification, I gained a deeper understanding of relational database management, advanced query techniques, and the intricacies of data manipulation and analysis. This certification strengthened my ability to write complex SQL queries for data extraction, transformation, and aggregation, enabling me to efficiently handle large datasets and perform in-depth analyses.

In my work, I have applied these skills to conduct sophisticated data analyses, optimize database performance, and ensure data integrity. For example, in my project on Electric Vehicle Population Analysis, I utilized advanced SQL queries to extract and analyze data from multiple states, providing critical insights into market dynamics and consumer behavior. My proficiency in SQL has not only enhanced my ability to work with complex databases but also improved my decision-making processes by ensuring data-driven, accurate insights.

This expertise has become integral to my work ethics, emphasizing the importance of precision, efficiency, and thorough data management. My approach is always data-first, ensuring that every decision is backed by a solid foundation of actionable data.
Datacamp

Datacamp

Data Analyst Associate

Data Analyst Associate

As a Data Analysis Associate, I have developed comprehensive expertise in key data management and visualization tools, transforming raw data into actionable business insights. Through mastery of PostgreSQL, I've honed my ability to write complex queries, manage databases efficiently, and extract meaningful patterns from large datasets. My proficiency in Python enables me to automate data processing workflows and implement sophisticated analytical solutions, while my advanced Excel skills support detailed financial modeling and statistical analysis.
I've applied these technical capabilities to streamline data preprocessing workflows, ensuring data quality and consistency across projects. My experience with Power BI has enabled me to create compelling visual narratives from complex datasets, delivering clear, actionable insights to stakeholders at all organizational levels. This combination of technical skills and practical application demonstrates my commitment to data-driven decision-making and continuous improvement in analytical methodologies.
Codecademy

Codecademy

BI Dashboards with Power BI Course

BI Dashboards with Power BI Course

Through intensive study in Business Intelligence Dashboard development with Power BI, I have mastered the creation of dynamic, data-driven visualizations that transform complex data into compelling business narratives. My expertise encompasses the complete Power BI workflow, from data connection and transformation using Power Query to developing sophisticated DAX measures and creating interactive dashboards that drive strategic decision-making.
This specialized training has equipped me with the skills to design intuitive dashboards that effectively communicate key performance indicators, trend analyses, and business metrics to stakeholders. I apply best practices in data modelling, implement advanced filtering capabilities, and create user-centric designs that ensure accessibility and meaningful data interpretation across all organizational levels.

Contact

I'm eager to connect with potential employers.

Let's discuss how I can contribute to your team!

Email Me