Projects
A collection of products, tools and experiments.
Real projects from my public work: each begins with a concrete problem and reveals a different way of building.
MboaCook
A cooking companion built around what people already have: available ingredients, meal ideas, planning and shopping become one coherent journey.
Problem
Having food at home does not always make it easy to decide what to cook. The ingredients are there, but inspiration, time and organisation are often missing.
Solution
MboaCook starts with the pantry and turns everyday constraints into relevant recipe suggestions and actionable meal planning.
Refworkers
A referral platform that helps people find nearby professionals through recommendations, detailed profiles and direct contact.
Problem
Finding a reliable professional quickly remains difficult, especially outside major cities or when a breakdown requires an immediate local response.
Solution
Refworkers organises discovery by trade and location, then builds trust through verified accounts, contextual reviews and a right of reply for professionals.
VisionLit
A computer vision platform that turns images and documents into usable information through accessible services.
Problem
Detecting objects, reading codes or extracting information from documents often requires a processing chain that is difficult to integrate into a product.
Solution
VisionLit brings several computer vision capabilities together behind a platform approach, shortening the path from a raw image to usable data.
MathMate
An open-source mathematics utility designed to calculate, solve equations, visualise functions and support learning.
Problem
Calculation tools are often either too basic or designed for existing experts, without helping users connect results with understanding.
Solution
MathMate brings calculation, equation solving, visualisation and learning into one experience that can evolve with its community.
Digit Recognition
An open-source application that trains a neural network on handwritten digits, then lets people draw a digit and receive a prediction.
Problem
Machine learning stays abstract until data, training and inference are brought together in an experience people can manipulate.
Solution
The project connects MNIST training with a drawing interface, making the full path from example to model to prediction visible.