SIGT – Angola’s Integrated Tax Management Platform
Java | Spring | JavaScript | HTML | CSS | Oracle SQL | REST APIs | Postman | Selenium | Docker | OOP | MVC | Agile
The SIGT platform is the main government system for managing taxes in Angola, supporting taxpayers and revenue collection nationwide.
Contributed a key feature enabling small and mid-size businesses to securely file monthly tax returns, supporting over 10,000 taxpayers and $10M+ in annual revenue collection.
Contributed to the platform using the following technologies as part of a team: Stack: Java, Spring; Frontend: JavaScript, HTML, CSS; Backend: Java, Spring, REST APIs; Database: Oracle SQL; Testing and Debugging: Postman, Selenium; Containerization: Docker; Software Design Principles: OOP, MVC; Development Methodology: Agile.
SuperIntro - Relationship Simulator
TypeScript | JavaScript | React.js | Next.js | Node.js | MongoDB | LLMs | NLP | Azure OpenAI | LangChain | Vector Databases | Zustand | Tailwind CSS | ShadCN UI | Agile
SuperIntro is an intelligent networking platform that automates the professional connection lifecycle, from initial scouting to in-person meetups, using personality-driven AI simulations.
Contributed to the platform by implementing the Meetups feature end-to-end, redesigning core application layouts including the Matching Page UI, and optimizing user session management. Additionally, enhanced the AI-driven simulated conversation prompt feature to improve the engagement and natural flow of AI-generated interactions before real-world meetups occur.
Developed using the following technologies: AI/ML: LLMs, NLP, Azure OpenAI, LangChain, Vector Databases, Prompt Engineering; Frontend: React.js, Next.js, Zustand, Tailwind CSS, ShadCN UI; Backend: Node.js, TypeScript; Database: MongoDB; Tools & Methodology: Git, GitHub Copilot, Postman, Agile.
BuyHere E-Commerce Web App
JavaScript | React.js | Express.js | Node.js | MongoDB | JWT | Redux | RESTful APIs | PayPal REST API | Cloudinary | Tailwind CSS | Lucide | ShadCN UI | Jest | Postman | Render
BuyHere is a full-stack MERN e-commerce web app that allows users to authenticate, browse products, make purchases, view order history, track shipments, and review products, while allowing admins to manage inventory and orders processing.
Developed using the following technologies: Stack: MERN; Frontend: JavaScript, React.js, Redux, Lucide, ShadCN UI; Backend: JavaScript, Express.js, Node.js, JSON Web Token (JWT), RESTful APIs; Database (Cloud): MongoDB Atlas (NoSQL); Payment Method: PayPal REST API; Testing and Debugging: Jest, Postman; Images Storage Cloud Hosting Provider: Cloudinary; Web App Cloud Hosting Provider: Render.
Sleepify Hotel Booking Web App
Python | Django | JavaScript | jQuery | AJAX | HTML | CSS | Bootstrap | PostgreSQL | Supabase | JWT | RESTful APIs | Stripe API | PayPal API | Pytest | Postman | Render
Sleepify is a full-stack Django-based web app for hotel bookings that allows users to authenticate, browse hotels, book rooms, manage their profiles, and review hotels, while allowing admins to manage hotels, rooms, and bookings.
Developed using the following technologies: Stack: Python (Django), PostgreSQL; Frontend: Django, JavaScript, jQuery, AJAX, HTML, CSS, Bootstrap, Jazzmin; Backend: Python, Django, JSON Web Token (JWT), RESTful APIs; Database (Supabase): PostgreSQL; Payment Method: Stripe API, PayPal REST API; Testing and Debugging: Pytest, Postman; Database and Image Storage Cloud Hosting Provider: Render; Web App Cloud Hosting Provider: Render.
AI Tool Suite Web App
Python | Gemini API | Multimodal AI | LLMs | NLP | RAG | Vector Database | Embeddings | Chroma | LangChain | Streamlit Cloud
AI Tool Suite Web App is an AI-driven platform that enables users to interact with a conversational AI assistant, generate blog posts, and upload and analyze documents (PDF, CSV) and URLs, powered by artificial intelligence.
Developed using the following technologies: Frontend: Streamlit (Python); Backend: Python; AI/ML: Gemini API, LLMs, NLP, Multimodal AI, LangChain, RAG, Embeddings; Vector Database: Chroma; Cloud Hosting: Streamlit Cloud.
Bank Payment Fraud Detection
Python | Dash | Flask | Data Science | Machine Learning | XGBoost | Random Forest | KNN | Scikit-Learn | Pandas | NumPy | Plotly
Developed machine learning models to identify fraudulent bank transactions using the Banksim dataset, with nearly 600,000 simulated financial operation records.
The project involved data balancing techniques (SMOTE), exploratory analysis, and building predictive models such as KNN, Random Forest, and XGBoost. The XGBoost model stood out with an accuracy of 99.15%, precision, recall, and F1-Score of 0.99, as well as an ROC-AUC of 0.99.
The final product is a Python web app with Dash (Flask), showing exploratory analysis, model performance, confusion matrices, ROC curves, and feature importance charts. The web app provides actionable insights for fraud analysts and risk teams.
Bank Loan Default Prediction
Python | Dash | Flask | Data Science | Machine Learning | Random Forest | Gradient Boosting | Decision Tree | SVM | Logistic Regression | Scikit-Learn | Pandas | NumPy | Plotly
Developed machine learning models and data analysis techniques to predict the probability of loan defaults among bank customers, using a historical dataset consisting of customer information, accounts, transactions, and loans.
The project involved complex data preparation, feature engineering, and the evaluation of multiple models, including Random Forest, Decision Tree, Gradient Boosting, SVM, and Logistic Regression. Ensemble models like Random Forest and Gradient Boosting achieved exceptional results, with 100% Precision, Recall, F1-Score, and Accuracy.
The final web application, built with Python and Dash (Flask), features exploratory analysis, detailed model performance, confusion matrices, ROC curves, and variable importance graphs, providing actionable insights for bank managers, risk analysts, and customer service teams.
Telecommunications Customer Churn Prediction
Python | Dash | Flask | Data Science | Machine Learning | LightGBM | Random Forest | Gradient Boosting | Scikit-Learn | Pandas | NumPy | Plotly
Developed machine learning models to predict customer churn at a telecommunications company, using a historical dataset with over 3,000 records containing demographic information, subscribed plans, usage behavior, and payments.
The project involved data cleaning and preparation, feature engineering, and evaluation of multiple predictive models. The LightGBM model stood out, achieving an F1-Score of 0.88 and an AUC close to 0.95, enabling reliable identification of high-risk churn customers.
The final product is a Python web app with Dash (Flask), presenting exploratory analysis, model performance metrics, confusion matrices, ROC curves, and feature importance charts. The tool provides actionable insights for marketing, customer retention, and product management teams, helping implement proactive strategies to reduce churn.
Oil Field Production Data Analysis
Python | Flask | Dash | Data Science | Data Analysis | Pandas | Matplotlib | Seaborn | Openpyxl
Performed data analysis on oil well production from Volve’s oil field using a dataset of 15,000 records. The goal was to gain insights into well performance and assist oil management in decision-making.
The analysis identified the most productive wells (5599 and 5351) and those approaching the end of their economic life due to high water production, resulting in actionable recommendations to optimize resource allocation and operational decisions.
Developed using Python and libraries such as Pandas, Matplotlib, Seaborn, and Plotly.
Gender Recognition from Facial Images with Deep Learning
Python | Machine Learning | Convolutional Neural Network | Classification | Keras | Tensorflow | Numpy | Pandas | Matplotlib | Seaborn | Jupyter Notebooks
This project was presented at the 2024 Kansas Capitol Graduate Research Summit, earning a prestigious acknowledgment from the state of Kansas, the Kansas Governor, policymakers, and Fort Hays State University for its contribution to research and innovation.
Developed a convolutional neural network to classify a person's gender from facial images, achieving 91% validation accuracy using a dataset of over 23,000 images.
Potential applications include demographic analysis, targeted advertising, security enhancements, personalized user experiences, and medical diagnostics.
AES and RSA Encryption
Python | Streamlit Cloud | Visual Studio Code
This Python application encrypts and decrypts messages, supporting AES and RSA ciphers.
Developed the encryption and decryption algorithms from scratch, creating a system that shows each step of the AES process, including the initial key, round keys, and ciphertext for each 16-byte block.
The AES cipher uses a key size of 256 bits, and the RSA cipher can generate public key pairs (n, e) and private key pairs (n, d) with 80 or 128 bits.
Network Design and Implementation
Cisco Routers and Switches | PuTTY | Microsoft Visio
Designed and implemented a network project for a company.
Equipment: Cisco routers and switches, servers, and computers.
Configured the following services on Cisco routers and switches: GRE VPN, OSPF, hostnames, encrypted passwords, interfaces, clock rate, login, spanning tree, VLANs, ACL, GRE Tunnel, NAT, DNS, PPPoE, and ISP2.
System Administration using Linux
Linux (Red Hat Enterprise Server, CentOS, Ubuntu) | Windows | VMware Workstation
Configured and administered server and network services on Red Hat Enterprise Linux for a company, adapted to its business volume.
Configured essential services such as IP, DHCP, DNS, HTTP, FTP, NIS, NFSv4, YUM, Samba, OpenLDAP, LVM setups (PV, VG, LV), SWAP memories on two Red Hat Enterprise Linux servers, and configured DHCP in CentOS Linux clients.
Enhanced security by restricting server access to admin accounts, securing department folders, and using NFS to share directories between servers and clients.