Open, ethical AI and software development that puts privacy and ecology first.
We try to conduct our research as openly as possible. We avoid ‘black boxes’ wherever we can: to make progress, we want the community to be able to examine, understand and build on existing work.
To avoid approaches that simply wrap third-party services, our projects are designed to minimise external dependencies and keep the technical architecture as transparent as possible. Components are selected and vetted with a focus on autonomy, auditability and maintainability.
We avoid unnecessary artificial restrictions whenever possible, while keeping fair-use, security and service stability requirements in mind.
Our optimization-first and simple architectural design approach aims to deliver strong performance with lower energy consumption and hardware requirements.
We design our services to minimize data transfers and prioritize local or controlled processing whenever possible. We develop techniques without unnecessary centralized data collection.
lightweight & open AI content detector
ZoryaTrace is a project designed to analyse texts and estimate whether content may be AI-generated, with methods adapted to each case to improve efficiency.
lightweight & modulable flashcard app
Create, import, and study your own flashcard decks with a simple JSON-based system, efficiently - while the app is designed to be as lightweight as possible, your data is never uploaded and stay safe on your device.
Intelligent file sorting assistant
Private, secure, low-resource file sorting assistant powered by a self-learning local algorithm.
Doc Finder
Don’t worry, it’s not a(nother) dumb chatbot ;) Enter a keyword or a short query; the algorithm will find the most relevant documentation for you. The information may sometimes be incorrect.
> Try: “detector”, “flashcards”, “privacy”, “README”.