Products, tools, and experiments I build around AI, voice, desktop context, developer workflows, personal knowledge, and local-first software.
Most of my projects start from friction.
Something feels too slow. Too fragmented. Too dependent on cloud tools. Too far away from how I actually want to work. So I try to build a better layer for it.
Contents
- Main Product
- NovaVoice
- AI & Knowledge Tools
- Vault Chat Agent
- AI Bookmarks
- Desktop & Context
- Basic Context Resolver
- Gimme Transcript
- Infrastructure
- Helm Charts & Self-Hosted Tools
- Older Experiments
- What connects all of this
Main Product
NovaVoice
![NovaVoice cover — desktop voice OS, generated project cover]
A desktop voice OS for working with tools through voice.
NovaVoice is my main product focus.
The idea is not to open another AI chat window and type one more prompt. The idea is to build an operating layer where voice, desktop context, AI agents, and real tool integrations work together.
I want voice to become a practical interface for everyday work: writing, searching, navigating apps, managing email, working with calendars, controlling desktop actions, and reducing the distance between intention and execution.
What I’m exploring:
- voice as a real interface
- desktop context awareness
- AI agents connected to actual tools
- MCP-style integrations
- local desktop workflows
- faster interaction between human intent and software action
NovaVoice is where most of my current thinking about AI, interfaces, automation, and product design comes together.
Status: active product
Category: voice OS, AI agents, desktop automation
Links: NovaVoice website
AI & Knowledge Tools
Vault Chat Agent
![Vault Chat Agent cover — Obsidian AI assistant, generated project cover]
An AI assistant plugin for Obsidian vault chat, search, and reviewed edits.
Vault Chat Agent is an experiment around AI inside a personal knowledge base.
I’m interested in AI that helps with notes carefully: search, context, conversation, graph-aware retrieval, and reviewed edits — not uncontrolled rewriting.
The project explores a more controlled way to work with long-term notes: the assistant can help investigate, summarize, propose edits, and work with context, but the final decision stays with the user.
What it explores:
- AI over personal notes
- Obsidian vault search
- graph-aware retrieval
- reviewed edit proposals
- long-term personal context
- local-first thinking environments
Status: open-source experiment
Category: personal knowledge, Obsidian, AI tooling
Links: GitHub / Obsidian plugin page
AI Bookmarks
![AI Bookmarks cover — browser knowledge layer, generated project cover]
A Chrome extension for AI-assisted bookmark management.
Bookmarks usually become a graveyard of links. Saved, forgotten, never opened again.
AI Bookmarks is an attempt to make browser memory more useful: categorize saved pages, organize links, and turn scattered bookmarks into something closer to a lightweight personal knowledge layer.
The larger idea is simple: saved web content should be searchable, structured, and alive — not buried in folders forever.
What it explores:
- AI-assisted bookmark categorization
- browser-based knowledge organization
- OpenAI-compatible endpoints
- lightweight retrieval over saved links
- reducing friction in personal research workflows
Status: early experiment
Category: browser extension, AI, personal knowledge
Links: GitHub, Chrome Store
Desktop & Context
Basic Context Resolver
![Basic Context Resolver cover — desktop context layer, generated project cover]
A cross-platform desktop context resolver.
Basic Context Resolver detects active applications, browser URLs, and UI state across macOS, Windows, and Linux.
This is one of the building blocks behind a larger idea: AI tools become much more useful when they understand what is happening on your desktop.
A good assistant should not only wait for text input. It should understand the current app, active window, browser page, and the environment where the user is actually working.
What it explores:
- active app detection
- browser URL detection
- desktop UI context
- cross-platform context collection
- context-aware AI assistants
Status: open-source utility
Category: desktop automation, context, AI infrastructure
Links: GitHub / npm
Gimme Transcript
![Gimme Transcript screenshot — local transcription app]
Local desktop transcription for multi-speaker audio.
Gimme Transcript is a local transcription app for multi-speaker audio.
It is built around a simple principle: no cloud, no API keys, no data leaving your machine.
The app lets you import audio files, transcribe locally with Whisper models, label speakers manually per segment or in bulk, and export clean transcripts as Markdown or plain text.
What it explores:
- local transcription
- multi-speaker transcript workflows
- privacy-first desktop software
- local Whisper models
- simple exportable transcripts
- voice and audio as practical work inputs
Status: MVP
Category: desktop app, transcription, local-first
Links: GitHub
Infrastructure
Helm Charts & Self-Hosted Tools
![Infrastructure cover — deployment and self-hosted systems, generated project cover]
I also build smaller infrastructure and developer tools: Helm charts, deployment experiments, self-hosted utilities, scripts, and small pieces of backend/frontend glue.
These projects are not always polished products. Some are just practical tools made to solve a specific problem.
But they reflect something I care about: understanding the full path from idea to running software.
Frontend is my main background, but I care about the whole system — interface, backend, infrastructure, deployment, and the way all of it feels to work with.
Examples:
- Listmonk Helm chart
- Mautic Helm chart
- self-hosted tooling
- deployment experiments
- developer environment utilities
Status: utilities and experiments
Category: infrastructure, Kubernetes, self-hosted tools
Links: GitHub
Older Experiments
My GitHub also contains older projects: hackathon work, small APIs, validators, converters, Windows tools, and archived experiments.
I keep them there because they show the path.
Not every project has to become a company.
Not every experiment has to become a product.
Some projects are just traces of learning, curiosity, and technical movement.
What connects all of this
Most of my work orbits the same questions:
- How can AI become useful without becoming noise?
- How should voice work as an interface?
- How can software understand user context better?
- How can developer tools reduce friction?
- How can personal knowledge systems become more practical?
- How can local-first tools make work feel safer and more direct?
- How can small tools make focused work easier?
I like building things that remove friction.
Not everything has to be big.
But everything should teach something.