I build software across the web, desktop, mobile and embedded devices. My projects range from a workout tracker that follows a session onto a watch to a Pico on my desk that shows how this portfolio is doing.


Launcher
A customizable Android home screen
I built Launcher as a native Android home screen in C# and .NET 11. It organizes apps into customizable drawers and tiles with app search, gesture navigation and Android widgets. Root/Today brings notes, RSS feeds, checklists, reminders and calendar tools into the home screen, while Conversations groups notifications by app and supports inline replies.
Themes, accent colours, haptics and optional layered animations let you shape the interface. An optional system panel adds quick controls across apps, and a Windows companion extends it to the desktop.


WaTenancy
WhatsApp messaging for teams
An ASP.NET Core and Blazor Server application that wraps wacli so companies can manage multiple WhatsApp numbers, team access, API keys and webhook endpoints within their own workspace.


Smoke Jumper
.NET and database sandboxes in Docker
I built a Windows tool for spinning up local test environments from a .NET project or a published build. You can run the application, restore a database copy, or bring both up together. Each workspace keeps its containers, ports, logs and backup history in one place.
For recovery drills, it restores MySQL or SQL Server backups into separate containers and reports checks such as backup checksums, restored data comparisons and HTTP smoke tests individually.
Project walkthrough and screenshots · Backup and recovery write-up


Pico Studio
A workspace for Raspberry Pi Pico
I built Pico Studio to bring device files, a Python editor, diagnostics and the USB console into one desktop app. It can listen to a running board as soon as it connects. You enter the REPL explicitly when you need file access.
The editor includes syntax checks and Pyflakes diagnostics. You can transfer folders and ZIP archives and opt into live monitoring for heap usage, filesystem space and clock frequency.


Kevlog Desk Display
Portfolio analytics on e-paper
A Raspberry Pi Pico 2 W pulls aggregate analytics from my portfolio and cycles through daily, weekly, monthly and all-time views on a 2.13-inch Waveshare e-paper panel. It shows browser views, visitor estimates and download starts.
The firmware uses partial refreshes for routine updates, remembers multiple Wi-Fi networks and shows the last reading as stale if it loses access to the API. A setup portal lets me configure the device from a phone.


Kevlog
My portfolio and writing studio
I built the site behind this profile as a place to document projects, share technical notes and publish interactive examples. It has a SQLite-backed CMS for editing articles and project galleries, searchable topics, and aggregate analytics that also power my e-paper display.


Forge
Strength training with a 3D muscle map
I designed Forge to keep workout logging close to the session, whether you have your phone or just your watch. It spans Android, Windows and the web, with a native Wear OS companion for logging sets and tracking rest timers.
- A 3D Strength Map shows progression by muscle group.
- An optional AI coach uses your own provider key and lets you review proposed changes before applying them.


K2.Python
OpenClaw for the terminal
K2.Python streams replies from a self-hosted OpenClaw gateway into the terminal and accepts text or voice input. Its tool layer lets the agent work in a local development environment by reading and editing files, running shell commands, using Git and making web requests.


OpenClaw Skills
Reusable tools for OpenClaw agents
A collection of installable agent skills, including a fitness data connector that retrieves metrics such as steps, distance and active minutes.


K2.Maui
OpenClaw voice and chat for Android
A .NET MAUI client for talking to a self-hosted OpenClaw agent from Android. I built a messaging-style interface with voice input, spoken replies and a persistent WebSocket connection that can run in the background.


Twenty4Hire
Voice interviews and job discovery
At Dynamic Data Technologies, I helped design and develop a recruitment platform with spoken interview simulations, adaptive follow-up questions and candidate reports. The platform also includes map-based job discovery, PayFast payments and scheduled background processing.


BlazorComponents
Reusable UI components for Blazor
A collection of Blazor components you can copy into a project and adapt directly.


ZettelAi
Desktop AI chat and searchable notes
Zettel combines AI chat and note-taking in a translucent desktop overlay. You can dictate text, switch AI providers and create, search or edit notes from the same workspace.


LinkedIn Profile Analyser
Analyse LinkedIn profiles in Chrome
A Chrome extension that reads the profile you have open and passes its text to a selected Google AI model for analysis. A sidebar lets you save and reuse questions across profiles.


BERT Fine-Tuning
Train a text classifier on a laptop
A hands-on fine-tuning example using trading-related text as the classification problem.


MauiVisionONNX
Live camera classification on Android
I adapted the .NET MAUI ONNX sample to classify images from the live camera using inference on the device.


Local Phi-3 inference
Run Phi-3 on the CPU with .NET
A .NET and ONNX experiment that runs Phi-3 on the CPU and explores what local language-model inference looks like without a dedicated GPU.


Local Whisper transcription
Run speech-to-text on your own machine
An experiment in using OpenAI's Whisper model to turn audio into text locally.
If something is broken, understand it. If it can be improved, improve it. If it does not exist, build it.



