/research/applied-ai

Applied AI

Models that turn biosignals into meaning — and the same toolset applied to automation and decision support.

01 / Scope

What we work on.

01
Biosignal models
Classification and regression on EEG and other physiological signals, with attention to differences between subjects.
02
Time-series learning
Feature-based and deep learning methods for noisy, non-stationary sensor data.
03
Decision support
Systems that show a model's output together with its uncertainty, so that a person stays in the loop.
04
Automation
The same methods applied to operational data: classification, information extraction and workflow automation.

02 / Approach

Step 01
Evaluation before architecture
Metrics and data splits — subject-wise where it matters — are fixed before a model is chosen.
Step 02
Uncertainty is part of the answer
A prediction without a sense of how reliable it is is not useful for decisions.
Step 03
Baselines first
Every complex model is compared against a simple one. If it does not beat the baseline, it does not ship.
Step 04
From notebook to service
Models are packaged as versioned services with monitoring, not left in a notebook.

We work with research groups and companies on modelling problems where the data is noisy and the stakes are real.