Projects

Career Guidance system to bridge the guidance gap for secondary students, using peer clustering, semantic interest matching, and real-time economic indicators, to transform decisions from guesswork into data-driven strategies.

Directed a team to modernize radiation safety hardware, refactoring legacy code for maintainability and scalability. Enabled IoT capabilities via Bluetooth for real-time monitoring of healthcare worker safety data.

Secure donation platform integrating MPESA Daraja API for real-time payment validation and fraud prevention. Managed data persistence with Firebase for transparent, audit-ready records.

Secure messaging system using AES symmetric encryption with OTP-based key delivery via SMS simulation, enabling confidential communication between remote devices.

OCR-driven document management suite that converts scanned PDFs into searchable files using Tesseract and Ghostscript, with chunked processing for efficiency and complete local data privacy.

Trip planning tool that calculates fuel consumption and travel cost using real-time distance data from the OpenRouteService API, with Excel export and activity tracking.

Time series forecasting pipeline combining statistical models (Holt-Winters, SARIMA) and LSTM neural networks to analyze and predict stock price trends.

Fast-paced 2D arcade platformer in Unity featuring multiple gameplay mechanics, procedural systems, and a narrative-driven progression inspired by computer hardware restoration.

Multi-model machine learning API with Flask, integrating classification, clustering, and recommendation systems, and deployed it using Docker, Gunicorn, and NGINX for scalable, production-ready access.

NLP pipeline to extract topics and sentiments from student course evaluations, with an interactive Gradio interface deployed on Hugging Face for real-time predictions.

Low-cost IoT system for real-time air quality monitoring using environmental sensors, and an interactive dashboard powered by Home Assistant and Grafana.

Diabetes risk prediction system using an optimized MLP, enhanced with Explainable AI (SHAP, LIME), and deployed through a sustainable MLOps pipeline with monitoring and drift detection.

Performed comprehensive exploratory data analysis and multiple linear regression to model revenue drivers, identifying order volume and average order value as key predictors with strong statistical significance.

Machine learning model to detect fraudulent accounts for secure batch account validation in high-volume payment processing. Performed feature engineering and data exploration on fintech datasets to identify malicious activity.

Web-based analytical game that transforms Spotify playlist data into interactive gameplay