PROJECTS / MATHISEN DYNAMICS
From a concrete product to the technology behind it. Explore our development work, the purpose of each project and how we take it forward.
01 / LISA ACCOUNTING
Lisa is our accounting assistant in development. She helps extract information from documents and highlights what needs checking.
The initial focus is practical document work: receipts, accounting documents and the information they contain. Users should be able to see the source, correct information and follow up on uncertainties.
We develop features gradually and test concrete workflows. Information, review points and user experience are assessed together, so AI support makes the work easier to follow.
Ask us about Lisa ↗FROM DOCUMENTS TO CLARITY
These documents may be duplicates.
AI helps you find information. You check it and decide what is correct.
Illustration using sample data. Lisa is in development.
02 / ZILNO CORE
Zilno Core is the infrastructure connecting our AI agents to the models they use.
A shared foundation makes it easier to develop multiple agents and products without starting over each time. Core is the infrastructure around the models; it is separate from our language-model work in ZLM.
We extend the platform in focused steps, test how its parts work together and draw on internal projects such as Aria. The goal is a clear, flexible and controllable base for further development.
A SHARED CONNECTION
03 / ZLM — ZILNO LANGUAGE MODELS
ZLM is the project in which we develop and train our own language models.
We begin with focused tasks and controlled experiments. This lets us investigate what models learn, how they handle new examples and which changes actually help.
We develop, measure and compare step by step. Training and evaluation go hand in hand, with an emphasis on understanding results before moving forward. ZLM is a research and development project, not a finished general-purpose AI service.
Understand the result.
Build on what we learn.
04 / OTHER DEVELOPMENT AREAS
We explore how unwanted behaviour and anomalies in AI systems can be detected and investigated. The aim is better visibility and a clearer basis for follow-up.
The work is at an early stage, focused on bounded investigations and thorough evaluation.
Our view on security ↗Aria gives us a place to explore how an AI agent can work in everyday tasks, bringing together conversation, voice and practical tools.
We use these experiences to improve collaboration between people and AI.
Let Aria introduce herself ↗LET’S WORK IT OUT
Get in touch about Lisa, Core, our language models or opportunities to collaborate.
We help with everything from practical advice and AI support to larger digital development projects. Get in touch so we can explore your needs.
post@mathisen-dynamics.no